What’s new in CiteTrack AI

• Added optional OpenRouter for task drafts, Optimize fixes and helper-based analysis.
• Added model selection, encrypted key storage, save/test and disconnect.
• Added request-cost tracking and clearer connection errors.
• Preserved direct-provider connections; visibility probes do not use OpenRouter.
• Engine settings save/test inline without navigating away.
• Updated Grok’s search integration and clarified HTTP 410 errors.
• Gemini tests now require usable search-grounded answers.
• Failed or unverified checks no longer count as absent.
• Added per-engine visibility comparison with shared prompts, coverage counts and honest gaps for missing data.
• Added approved business facts with supporting URLs, review dates and revision history.
• AI claim checks compare saved answers against those approved facts.
• Added quote validation, repeat-check caching and historical-result labeling.
• Possible contradictions create evidence-backed review tasks without editing public content.
• Added AI-generated paragraph correction proposals with exact before/after review.
• Added explicit Apply confirmation, recovery copies and targeted Undo.
• Correct pages can receive “no change needed.”
• Added separate public-page verification and charge-confirmed background AI rechecks.
• Added live progress and previous/new answer comparisons.
• Unsupported builders and ambiguous targets remain manual.

• Added robots-directive contradictions, repeated response-header rules and same-language hreflang conflicts.
• Added duplicate primary schema-entity detection.
• Detected repeated exact GTM, GA4, Meta Pixel and Clarity installations.
• Grouped related findings into one task while protecting legitimate configurations from false positives.
• Combined page/AEO/GEO and internal-link checks under Refresh website checks.
• Kept progress active until both matching scans finished.
• Automatically completed stale link tasks only after sufficiently complete, verified checks.
• Updated task counts and cards after completion.
• Fixed a Technical Health loading error in the safe-fix panel.
• Added a dedicated latest-scan results screen with status filters.
• Retained bounded report membership after completion.
• Added protected CSV downloads and browser print/save-PDF.
• Clarified partial results and export limits.
• Preserved completion information beyond the previous transient lifetime and explained background scheduling.
• Fixed selected-button contrast and strengthened shared button states.
• Added issue-type and page/link search filters.
• Preserved filter selections through counts, pagination, refresh and CSV export.
• Simplified instructions, collapsed advanced details and added mobile result cards.
• Clarified unconfirmed scan status without falsely claiming completion.
• Added open-task trends with new, reopened and verified-resolution counts.
• Kept manual closures and deleted records separate from verified fixes.
• Retained up to 48 completed-refresh snapshots without inventing earlier history.
• Added direct reconnection and current-page selection for older tasks.
• Restored eligible Generate controls and archived old-page proposals during reconnection.
• Packaged 94 production files in a flat ZIP, with no bundled tests.

• Added Keep change and close beside Undo for verified content and SEO changes.
• Allowed users to stop pending measurement intentionally.
• Preserved live values while removing only CiteTrack’s temporary rollback information.
• Corrected Undo messages and removed unused renderer parameters and an unused hook.
• Added Keep and Undo for supported Rank Math Article and FAQ schema changes.
• Protected newer provider edits from being overwritten by Undo.
• Displayed the exact visible FAQ questions and answers for manual workflows.
• Added clearer page links and stale-evidence guidance.
• Added checks for schema that is correctly formatted but describes the wrong page type.
• Distinguished current-page entities from legitimate related articles or products.
• Blocked Article schema generation and writing for inappropriate page profiles.
• Consolidated overlapping mismatch findings into one guided task.
• Added explicit Article, Page/landing and Product classifications.
• Required complete schema to identify the current page—not a related URL.
• Added an explicitly approved CiteTrack-owned JSON-LD fallback when a safe native writer was unavailable.
• Required verified commerce facts for Product markup.
• Added duplicate/conflict checks, bounded storage, exact Undo and permanent removal controls.
• Detected repeated canonical, title, description and supported social tags from rendered HTML.
• Covered custom plugins, themes and snippets—not just recognized SEO plugins.
• Distinguished identical duplicate canonicals from conflicting destinations.
• Preserved valid repeated tags and avoided automatically deleting unknown output.

• Shows the exact linked words, nearest page heading, occurrence count and bounded surrounding text for every broken-link or redirect task.
• Stops treating non-web URI schemes such as about:, blob: and ftp: as same-site paths.
• Never labels a broken URL as a verified destination; only a confirmed redirect displays that label.
• Recognizes stale link findings before offering a replacement and provides one check that safely clears the task without editing the page.
• Separated verified problems from optional content suggestions.
• Stopped creating tasks or lowering scores based only on word count, age, lists, citation counts and similar heuristics.
• Retired older heuristic-only tasks while preserving their history.
• Restored complete page-task evidence and clearer “Needs action” versus “Optional” labels.
• Added evidence-based direct-answer and key-takeaway tasks.
• Required sufficient impressions and conservative search-performance signals before recommending changes.
• Used matching search queries and page facts for generated proposals.
• Batched Search Console requests and preserved existing findings when Google data was unavailable or incomplete.
• Consolidated qualification logic and removed a duplicate solution label.
• Added follow-up measurement for reviewed direct-answer and key-takeaway improvements.
• Saved the baseline and waited for a complete 28-day post-change period.
• Compared clicks, impressions, CTR and position without claiming the edit caused the result.
• Added bounded retries, dedicated scheduling and cancellation when changes were undone or tasks closed.

• Added exact same-site redirect updates to the guided Technical Health Safe Fix Queue, with complete old/final URL previews and one verifiable Undo per page.
• Broken, outside, fragment-bearing, stale or changed links remain explained decisions; CiteTrack does not guess a destination or silently remove a section anchor.
• Rechecks signed evidence, editor permission, page state, occurrence count and a direct working destination immediately before each write.
• Verifies the Undo backup before editing and rolls the page back if its task state cannot be saved.
• Detects and displays technical-scan save or cron failures instead of leaving an apparently frozen running scan.
• Correctly recognizes same-host HTTP-to-HTTPS redirects as internal while retaining existing scan and queue limits.
• Added a guided Safe Fix Queue that separates deterministic reversible fixes from tasks needing a decision or manual editing.
• Limits background batches to ten reviewed changes.
• Rechecks the signed preview, permission, page state, evidence freshness, durable destination and exact Undo boundary before every write.
• Applies only direct answers, key takeaways and visible buyer Q&A derived from existing page text.
• Metadata, links, schema and technical decisions remain individual workflows.
• Shows live background progress, item-level results and a direct Review or Undo path without requiring the Tasks screen to stay open.
• Fixed internal-link relevance scoring so it compares actual words instead of token-frequency numbers.
• Recognizes equivalent absolute and site-relative links before suggesting or inserting them.
• Prevents duplicate links to the same page.
• Made every one-click internal-link suggestion provably insertable into the current saved page.
• Shows the exact surrounding readable page text before inserting a link.
• Separates manual link ideas from ready-to-insert links.
• Prevents automatic links inside scripts, styles, code samples, form controls, existing links and shortcode attributes.
• Revalidates the selected destination and server-generated anchor at click time instead of trusting submitted form text.
• Added guarded Undo that removes only the exact link CiteTrack inserted.
• Undo refuses to overwrite subsequent page edits.
• Avoids running front-end content filters merely to rank internal-link candidates.
• Removed the redundant licensing validation request made before normal Google relay calls.
• Reuses current cached relay credentials.
• Expired, rejected and explicitly refreshed credentials continue to self-heal through license activation.
• Separated GA4 reporting from optional GA4 event sending in plain language.
• Search Console and GA4 reports require only the Google connection and selected property.
• Added an explicit GA4 event-sending switch.
• Added strict G- Measurement ID validation.
• Added protected secret fields and a remove-secret control.
• Added clearer ready, incomplete and off status indicators.
• Reporting-only saves ignore irrelevant browser autofill.
• Preserves the selected GA4 property when Google temporarily cannot list properties.
• Stops loading GA4 event delivery on public requests when optional event sending is disabled or incomplete.
• Stops scheduling daily Microsoft Clarity fetches without a Data Export token.
• Clears stale no-operation Clarity schedules.
• Avoids reading the Clarity secret during ordinary public requests.
• Removed obsolete Google-license compatibility branches.
• Removed the optional ctype dependency from relay-token expiry checks.
• Added WordPress regression coverage for relay-token reuse, GA4 configuration states and Clarity scheduling.
• No database migration was introduced.

• Removed the redundant licensing validation request made before normal Google relay calls.
• Reuses current cached relay credentials.
• Expired, rejected and explicitly refreshed credentials continue to self-heal through license activation.
• Separated GA4 reporting from optional GA4 event sending in plain language.
• Search Console and GA4 reports require only the Google connection and selected property.
• Added an explicit GA4 event-sending switch.
• Added strict G- Measurement ID validation.
• Added protected secret fields and a remove-secret control.
• Added clearer ready, incomplete and off status indicators.
• Reporting-only saves ignore irrelevant browser autofill.
• Preserves the selected GA4 property when Google temporarily cannot list properties.
• Stops loading GA4 event delivery on public requests when optional event sending is disabled or incomplete.
• Stops scheduling daily Microsoft Clarity fetches without a Data Export token.
• Clears stale no-operation Clarity schedules.
• Avoids reading the Clarity secret during ordinary public requests.
• Removed obsolete Google-license compatibility branches.
• Removed the optional ctype dependency from relay-token expiry checks.
• Added WordPress regression coverage for relay-token reuse, GA4 configuration states and Clarity scheduling.
• No database migration was introduced.
• Fixed Google Search Console connection failures caused by a cached license token that was present but expired or rejected.
• Invalid relay credentials now self-heal through an idempotent license refresh before Google OAuth begins.
• Google connection failures return users to Integrations with a useful administrator message instead of exposing a raw relay error page.
• Replaced the generic canonical URL editor with a plain decision:
  • Keep the current preferred URL.
  • Use the current page as the primary URL.
• Detects missing, alternate, invalid and multiple rendered canonical tags.
• Carries exact, bounded canonical evidence into Tasks.
• Allows one-click canonical changes only through one proven writable SEO metadata owner.
• Conflicting, generated and unknown canonical owners remain manual.
• Rechecks stale evidence before writing.
• Blocks automatic repairs when multiple canonical tags exist.
• Verifies the live result after writing.
• Rolls back failed verification.
• Retains guarded Undo.
• Prevents Undo from overwriting newer editor changes.
• Requires explicit completion before removing the temporary recovery record.
• Removes the older generic canonical editor and its unguarded Undo route.
• Keeps historical canonical recovery evidence visible for manual review.
• Prevents malformed or conflicting canonicals from being accepted as healthy.
• Preserves recovery evidence when rollback cannot be proven.
• Added exact public-page indexability checks for:
  • HTML robots meta tags.
  • X-Robots-Tag response headers.
• Added a plain-language decision workflow:
  • Keep an intentionally hidden page out of active work.
  • Remove a proven per-page noindex setting.
• Automatic removal requires one unambiguous SEO metadata owner.
• Global, theme, server-header and multi-plugin cases remain manual.
• Preserves unrelated robots directives.
• Stores an exact backup before changing indexability.
• Verifies the live page immediately after applying a change.
• Rolls back failed verification.
• Added guarded Undo.
• Updated rendered-page caching to retain bounded response-header evidence.
• Older HTML-only cache entries are treated as unknown instead of receiving a false pass.
• Added an optional “Also check links to other websites” setting for manual Technical Health scans.
• Automatic weekly scans remain internal-link only.
• Reuses the same exact-hop, safe-request and false-positive protections for outside links.
• Blocked, rate-limited, timed-out and temporary 5xx responses remain unverified instead of being labelled broken.
• Distinguishes internal and outside link tasks in plain language.
• Retains the verified replacement and guarded Undo workflow.
• Keeps outside-link scanning explicitly opt-in.
• Outside scans remain subject to existing page, link, batch, time and finding limits.

• Added a bounded Technical Health scanner for saved internal page links.
• Runs scans in small WP-Cron background batches.
• Shows live scanning progress.
• Does not run link scanning during public page views.
• Added exact broken-link and redirect-hop evidence.
• Timeouts, rate limits and access-denied responses are reported as unverified instead of being mislabelled as broken.
• Added a dedicated Technical Health task area.
• Added plain-language explanations and next steps.
• Deduplicates findings by page and destination.
• Added conservative one-click link repair.
• Validates the replacement before changing content.
• Updates only exact matching saved anchor links.
• Verifies the result after the change.
• Keeps one guarded Undo backup.
• Added a “Keep change and complete” checkpoint.
• Undo refuses to overwrite a page changed after the CiteTrack repair.
• Added weekly automatic Technical Health checks through a lightweight daily due check.
• Added manual rescans.
• Added lost-cron recovery.
• Added 30-day cleanup for unused Undo backups.
• Renamed crawler diagnostics from “Redirect chains” to “Redirect responses,” because one 3xx response does not prove a complete redirect chain.
• Packages production files directly at the ZIP root.
• Opening the archive now immediately shows citetrackai.php, includes, assets and the other plugin files.
• Keeps version numbers in release ZIP filenames.
• Preserves CiteTrack’s verified in-place replacement flow for existing installations.
• Updated the production package gate to reject an unnecessary citetrackai/ wrapper directory.
• Rewrote task headings and explanations in plain language.
• Added clear “What this means” and “What to do now” sections for every task family.
• Moved optional scan scores and implementation evidence under collapsed Technical details.
• Keeps exact AI answers and source evidence easy to inspect.
• Replaced raw evidence field names with readable labels.
• Moved optional progress administration into a closed accordion.
• Added direct, provider-aware instructions for schema tasks.
• Users no longer need to understand schema ownership or reversible-write terminology to determine the next step.

• Repairs historical page tasks that retained a local affected URL but lost their WordPress post identity.
• Restores Generate, Apply and Recheck after the page is proven to be a local editable WordPress page.
• Pauses generation and writes when an older task URL cannot be mapped to a real editable page.
• Provides one clear instruction to refresh website checks when mapping cannot be confirmed.
• Removed the contradictory state where a task showed Generate and Apply controls but the secure handler could only reject the operation as unavailable.
• Aligned the live meta-description quality indicator with the stated 105–160-character writing target.
• Keeps release ZIP filenames versioned while mapping a verified uploaded CiteTrack package to the folder of the copy currently running.
• WordPress now offers an in-place replacement instead of installing another CiteTrack copy.
• Makes self-hosted update checks, integrity verification and reactivation use the actual installed plugin basename.
• Supports installations residing in older version-numbered plugin folders.
• Refuses to rename unverified uploaded packages.
• Never deletes an unexpected temporary-folder collision during package preparation.
• Preserves WordPress’s version-comparison confirmation for manual uploads.
• Forced overwrite remains restricted to authenticated CiteTrack update operations.
• Added an Optimize readiness card showing the detected owner for:
  • Meta descriptions.
  • SEO titles.
  • Canonical URLs.
• Stops creating SEO-title work solely because a title exceeds 60 characters.
• SEO-title tasks now require a genuinely missing published SEO title.
• Retires historical title-length-only tasks in bounded batches while preserving them as completed history.
• Uses available Search Console queries to inform missing-title generation.
• Rejects placeholder or excessively long AI-generated titles instead of silently cutting words off.
• Detects IndexPulse SEO Lite as the active per-page SEO metadata owner.
• Stops incorrectly asking IndexPulse users to install another SEO plugin.
• Saves reviewed meta descriptions and SEO titles using IndexPulse’s actual two-record storage contract:
  • The custom value.
  • Its matching source selector.
• Updates and verifies both IndexPulse records together.
• Keeps an exact backup of both records.
• Undo restores the prior values or their prior absence.
• Treats IndexPulse’s permalink-derived canonical URL as detected but read-only.
• Prevents guessed post-meta canonical writes.
• Prevents duplicate canonical tags from being added to the frontend.
• Clarifies that CiteTrack supports registered custom metadata destinations.
• CiteTrack deliberately does not guess based on arbitrary fields merely named “title” or “description.”

• Makes the Tasks website-check status update automatically while the bounded background scan runs; no manual page reload is required to see checked and remaining pages.
• Adds a visible, accessible progress bar with scanning, finishing, reconnecting and completed states so temporary polling trouble does not look like a stopped job.
• Keeps scanning independent of the Tasks page, reduces polling while the tab is hidden and refreshes immediately when the user returns.
• Updates task-area counts when reconciliation finishes and offers one clear “Show updated tasks” action instead of silently leaving the displayed queue stale.
• Prevents duplicate refresh submissions while a scan is active and explicitly schedules a final task reconciliation after the last page is checked.
• Qualifies meta-description work before it reaches the active queue: historical character-count-only findings are retired in a bounded pass, while a future genuinely missing tag can reopen the same task.
• Adds one visible website-check refresh control to Tasks, shows refresh progress, and returns the user to the queue while the bounded background scan runs.
• Detects Search Console automatically and shows Not connected, Connected and last successful working states with a direct setup action when needed.
• Removes the extra “check whether a rewrite is justified” detour; existing descriptions require measured Search Console evidence before they can remain actionable.
• Stops completed or unsupported page findings from rendering editing controls, and rebuilds the task queue after fresh page evidence arrives.
• Preserves the selected crawler identity filter when switching the identity chart between Bar and Line.
• Adds clear vertical breathing room between the daily crawler-activity chart and the access/identity KPI cards.
• Gives the “Who is reading your site?” identity trend its own Bar and Line control while preserving the selected date, company, identity, search and Advanced filters.
• Uses stacked bars for an at-a-glance identity total and separate lines for comparing verified, user-agent-only and possible-spoof changes over time.
• Restores the stacked daily crawler-company chart directly below the shared Crawl Health filters, so each day shows which crawler networks visited and how many requests each made.
• Keeps day, 48-hour and 60-minute intervals plus bar/line controls available in the normal workflow instead of hiding them in Advanced diagnostics.
• Retains the separate identity-confidence trend and mix, while heavier response, page-value and crawl-waste investigations remain opt-in.
• Restores the day-by-day verified, user-agent-only and possible-spoof identity trend to the normal Crawl Health dashboard while keeping expensive crawl-waste and page-value diagnostics opt-in.
• Repairs orphan crawl rows during every DB20 recovery attempt even when an earlier partial migration already created the unique crawler-name index.
• Keeps exact answer and Share-of-Voice pagination while counting prompt/engine slots from the covering database index; a one-million-citation benchmark reduced this path from 14.2 seconds and 41 queries to 0.92 seconds and 2 queries.
• Adds regression coverage for the indexed partial-migration state and for preserving a visible SEO recovery control when the original metadata provider is unavailable during Undo.
• Ships only the production plugin under one citetrackai/ directory; tests and nested archives remain outside the installable ZIP.
• Stops treating a 120–160 character range as a Google pass/fail rule; a meta-description task is now created only when the tag is actually missing.
• Stops older length-only tasks from proposing a rewrite and directs the user to rescan and close them.
• Rebuilds generation around the page title, excerpt, content, saved focus phrase and connected Search Console queries instead of a generic score.
• Adds a cached 28-day Search Console page comparison with clicks, impressions, CTR, average position and top real queries; GA4 sessions/views are clearly labelled as post-click context.
• Rejects editor placeholders, empty/very short output and essay-length output, retries one invalid AI response, and never silently chops a sentence at 160 characters.
• Explains the exact recommendation basis, data availability and permanent SEO-plugin destination before the user applies anything.
• Turns missing image descriptions into a guided page workflow: review each Media Library image, save a meaningful description, or explicitly mark a decorative image.
• Writes reviewed image descriptions to both the WordPress Media Library and only the matching saved page tags that still lack an alt attribute; existing image descriptions are left untouched.
• Adds bounded exact backups, visible change history and per-image Undo so users can restore the precise prior Media Library value and page markup.
• Verifies both writes before reporting success and rolls the page back when WordPress cannot confirm the Media Library update.
• Keeps dynamic, theme and third-party images out of automatic writes when CiteTrack cannot identify a safe Media Library owner, while providing a clear manual path.
• Makes image-description changes visible in Optimize’s Applied fixes area without presenting a misleading free-form page-content editor.
• Gives every actionable failed Optimize check one plain “Fix this” next step that opens the evidence and safest available workflow for that exact page.
• Makes the guarded Article and FAQ schema workflows reachable directly from their failed Optimize rows instead of requiring users to discover them elsewhere.
• Renames the final table column to “Next step” and adds a short instruction above the checks so a first-time user knows where to begin.
• Shows Undo only when CiteTrack can prove it owns a complete marked-content boundary or an exact metadata/schema backup; naturally passing checks no longer expose a misleading Undo button.
• Stops reporting an Undo as successful when no exact prior state can be verified or restoration fails.
• Shows durable Key takeaways and provider-native Article/FAQ schema changes accurately in Applied fixes, with one Review path and an explicit explanation of where the change lives.
• Adds an editable one-click Key takeaways fix for article pages that have no scannable list or table.
• Builds the immediate suggestion only from sentences already saved on the page, while connected AI alternatives are instructed to use only the bounded page evidence.
• Requires at least two reviewed items, strips submitted markup and list prefixes, removes duplicates, caps the list at eight items and limits each item to 240 characters.
• Saves the approved list as ordinary WordPress blocks near the top of the article, immediately verifies the same AEO check and keeps the content active when CiteTrack is disabled.
• Replaces only CiteTrack’s marked Key takeaways block on later edits and gives Undo a precise boundary so unrelated page content remains untouched.
• Confirms WordPress stored the exact proposed content before reporting any marked content fix as applied, preventing filters or competing saves from producing a false success.
• Keeps WebPage schema guided: Rank Math normally publishes WebPage markup itself, so CiteTrack does not add a potentially duplicate native WebPage record.
• Compares the exact visible FAQ answer text with public acceptedAnswer.text instead of passing when only the question name and any non-empty answer are present.
• Extracts bounded question-and-answer pairs from details/summary FAQs, Rank Math FAQ blocks and CiteTrack’s durable buyer-FAQ section; unpaired questions are never offered for an automatic schema write.
• Adds a guarded one-click FAQPage fix when Rank Math is the single initialized schema owner and CiteTrack can prove the public page has no existing FAQPage markup.
• Shows every exact visible question and answer before approval and writes one non-primary FAQPage record to Rank Math native storage so existing Article schema can remain in place.
• Refuses the write when the public page cannot be verified, a Rank Math FAQ block or FAQPage record already owns the markup, an oversized schema record cannot be safely inspected, or an intervening record appears.
• Verifies the public FAQ after writing, keeps it active when CiteTrack is disabled, and supports exact Undo from a private backup.
• Verifies the complete public Article, WebPage, Product and FAQPage evidence needed by each check instead of treating a schema type name alone as a pass.
• Shows the exact missing or unmatched schema fields, valid types, invalid JSON-LD blocks and evidence source directly inside the task.
• Stops ordinary question headings and generic disclosure widgets from creating false FAQ schema tasks; only recognized visible FAQ structures are matched to public questions and answers.
• Keeps public JSON-LD inspection bounded to 500 nodes and reports when a larger graph was only partially inspected.
• Adds a guarded one-click Article schema fix when Rank Math is the single active, initialized schema owner and the page has no existing schema record.
• Writes the reviewed Article facts to Rank Math’s native storage, immediately rechecks the public page, remains active when CiteTrack is disabled, and supports exact undo from a private backup.
• Refuses automatic schema writes when another provider owns the page, multiple SEO plugins are active, Rank Math Schema is unavailable, or an existing/intervening schema record could be overwritten.
• Keeps the previous database version in place until the final DB20 schema, independent proof and cleared error state have all been stored successfully; any failed completion write revokes the proof and keeps setup paused.
• Keeps the previous database version active when any DB20 migration step, schema write or final validation fails, so an incomplete upgrade is never recorded as complete.
• Independently revalidates sites that an older release already stamped DB20, preventing a previously partial migration from being trusted merely because its version option says 20.
• Verifies every declared table, column, primary key, exact index column order and index uniqueness plus DB20 crawler identities before resuming tracking, and shows administrators a recoverable retry error when verification fails.
• Adds a destructive DB19-to-DB20 benchmark that forces missing-index, wrong-column-index and wrong-index-uniqueness failures, proves the version is preserved, then validates a successful retry while remapping 100,000 representative crawler rows.
• Replaces the Reputation and Share-of-Voice full-history grouping queries with bounded latest-result keyset pages.
• Loads Reputation prompts with their answer rows and fetches visible claim analyses in one batch, removing per-answer database lookups.
• Extends the scale benchmark to verify descending latest-answer correctness and the public Share-of-Voice pagination response.
• Preserves the exact original SEO metadata state on undo, including the difference between an absent provider field and one intentionally stored as empty.
• Tightens admin output escaping at the final render boundary and removes the internal-link candidate query’s exclusionary WordPress query parameters.
• Shows where each meta description, SEO title or canonical change will be saved before the user applies it.
• Writes reviewed SEO metadata to Yoast, Rank Math, All in One SEO, SEOPress, The SEO Framework or Slim SEO using each plugin’s own storage.
• Stops automatic metadata changes when multiple SEO plugins are active, avoiding an arbitrary destination and possible competing tags.
• Keeps proposals editable and copyable when no supported SEO plugin is active, with a clear path to add a permanent destination.
• Treats the SEO title separately from the visible WordPress page title, so an optimization no longer changes the on-page heading or navigation label.
• Backs up the original provider value, supports exact undo, and preserves undo compatibility for fixes applied by earlier CiteTrack versions.
• Rechecks the public page immediately after applying a fix and reports whether it passed, still differs, or could not yet be verified.

• Prevented internal page navigation from being counted as multiple referral acquisitions.
• Improved citation processing for accounts with large prompt histories.
• Made the Sources workspace load from the latest results instead of repeatedly scanning full history.
• Reduced crawler dashboard queries and unnecessary calculations.
• Improved database migrations for large datasets.
• Added performance verification for million-row crawler logs and large citation histories.
• Added a Storage screen showing database size, row counts, index usage, cleanup status, and manual cleanup controls.
• Added configurable data retention between 30 and 365 days.
• Preserved frozen reports and active workflow data during cleanup.
• Limited stored AI answers and citation evidence to practical amounts.
• Added bounded cleanup for expired rate-limit and abuse-protection records.
• Capped generated drafts, reputation history, and attribution snapshots.
• Removed four redundant database indexes.
• Improved crawler identity storage and concurrent crawler lookups.
• Added automatic repair of old draft and report storage backlogs.
• Prevented duplicate open incidents for the same alert rule.
• Automatically restores the daily cleanup schedule if WordPress loses the scheduled event.
• Limited inherited reputation email queues.
• Removes old, untouched automatic task suggestions while preserving assigned tasks, notes, drafts, outcomes, and source work.
• Preserves manually saved report baselines while limiting unnecessary automatic snapshots.
• Removes optimization scores when their WordPress page or post is permanently deleted.
• Repairs legacy score records belonging to deleted content.
• Recovers missing one-time Optimize scan events.
• Properly closes alert deliveries interrupted during their final attempt.
• Cancels pending notifications when an alert rule is deleted.
• Added limits to temporary sitemap storage, including URL count, length, size, and origin checks.
• Replaced expensive full-table maintenance counts with indexed existence checks.
• Preserves conversions connected to older referrals during data cleanup.
• Added database-backed ownership locks for cleanup, upgrades, Optimize scans, sitemaps, tasks, insights, and license jobs.
• Unified alert deletion so incidents and queued notifications are consistently cancelled.
• Prevented test alerts from remaining in the normal delivery queue.
• Repairs legacy test and parentless alert records.
• Removes orphaned alert deliveries and generated drafts in controlled batches.
• Reduced duplicate page-content storage during forced Optimize scans.
• Removes temporary sitemap state when the plugin is deactivated.
• Rebuilds historical citation insights in smaller batches to reduce database locking.
• Prevented an expired background worker from deleting the lock owned by a newer worker.
• Added safer shared locking for citation queues, batches, and alert evaluation.
• Prevented simultaneous reputation email jobs from overwriting each other.
• Added a persistent retry queue for reputation digest emails.
• Prevented overlapping reputation email sends.
• Moved recovery and maintenance checks away from normal public-page requests.
• Stopped loading Search Console administration code on public requests that only need GA4 attribution.
• Automatically recovers claimed alert deliveries and crawler-verification jobs when their lease expires.
• Reschedules alert retries when a worker starts before plugin settings are available.
• Makes sitemap scans recoverable after PHP failures, network timeouts, or interrupted duplicate events.
• Builds large citation runs in controlled 500-prompt pages.
• Avoids loading an account’s complete prompt and citation history into memory.
• Uses optimized indexes to locate the latest prompt runs.
• Streams citation queue records in small batches for improved memory and database performance.

• Prevented overlapping visibility checks from charging for or saving the same prompt and AI service twice.
• Added complete WordPress multisite support for activation, deactivation, new-site setup, and recovery.
• Pending alerts and reputation emails now resume automatically after reactivation.
• Removed reporting limits that could hide results on accounts with thousands of prompt and AI-service combinations.
• Fixed MCP batch errors and improved rate-limit enforcement.
• MCP access is now revoked when its WordPress administrator is removed or loses administrator privileges.
• Protected the plugin from unusually large API, webhook, update, and robots.txt responses consuming excessive memory.
• Improved the updater so the plugin is reactivated only when it was active before the update.
• Full-data uninstall now removes remaining author-profile fields, notices, and rendered-page caches.
• Protected CSV exports from spreadsheet formula injection.
• Clarified that crawler IP addresses are always privacy-hashed and verification data is encrypted and temporary.
• Corrected privacy disclosures for licensing, updates, Google Search Console, and user-requested AI calls.
• Improved protection when downloading official AI crawler network information.
• Added safer database locking and accurate progress totals for scheduled visibility checks.
• Updated cost estimates for newer OpenAI, Gemini, and Grok models.
• Added regression tests for privacy, crawler-network downloads, queue locking, usage accounting, and provider pricing.
• Prevented unrelated domains—especially country-code domains such as .co.uk sites—from being mistaken for the tracked business or a competitor.
• Improved detection of genuine AI referral traffic while rejecting misleading or lookalike domains.
• Advanced alerts now correctly follow the selected page, crawler, prompt, AI service, and other filters.
• Response-time alerts now use the complete selected period instead of sampling only recent records.
• Added expanded domain, attribution, alert, and WordPress compatibility testing.
• Limited spoofed crawler traffic so fake bot names cannot grow logs and verification queues indefinitely.
• Reduced unnecessary database work on every crawler request.
• Improved public rate limiting with a single atomic database operation.
• Prevented duplicate visitor-attribution beacons across browser tabs while preserving legitimate campaign changes.
• Improved Crawl Health performance and moved expensive diagnostics behind the Advanced Diagnostics option.
• Tested Crawl Health with 500,000 records while keeping the normal screen under five seconds.
• Removed confirmed obsolete code without affecting the Cost dashboard or existing features.
• Added crawler-burst, attribution-beacon, and production-package regression tests.
• Kept development tests out of the installable plugin ZIP.

• Removed an obsolete private reputation-email method. The active batched email digest and notification setting remain unchanged.
• Moved bot-catalog checks, encrypted-vault migrations, bot-queue recovery, and alert-queue recovery off ordinary public page views. They still run safely through WordPress administration and WP-Cron.
• Reduced plugin-added database queries on ordinary public requests from six to three.
• Removed a stale internal README that no longer represented the plugin.
• Excluded development-only regression tests from the installable ZIP while retaining them in the development source.
• Completed a callback-aware review of all 422 private methods and found no remaining confirmed dead methods.
• Retained active callback-only code, legacy scheduled hooks, AI provider integrations, public helpers, and chart dependencies.
• Reduced the installable ZIP by 7,036 bytes compared with 1.79.6.
• Passed PHP, JavaScript, security, Analytics, session-token, WordPress runtime, WP-Cron, packaging, and crawler-performance tests.

• Adds custom From/To date controls to Analytics, matching the AI Visibility range picker.
• Preserves the selected range when switching Analytics tabs.
• Keeps quick presets for 7 days, 30 days, 90 days, 12 months, month to date and last month.
• Updates the AI lens in Analytics so its totals and chart labels follow the selected reporting window.
• Adds Google-Analytics-style trend charts above the Analytics tables, including traffic, revenue, channel trends, channel share, top sources, source trends, top landing pages and AI share by landing page.
• Keeps the existing detail tables below each chart so users can visually spot movement first, then audit the exact rows.
• Adds Cloudflare AI Crawl Control crawler names to the local crawler catalog, including Anchor Browser, Cloudflare Crawler, Manus Bot, MistralAI-User, Novellum AI Crawl, ProRataInc, Terracotta Bot, TikTok Spider, archive.org_bot and Arquivo Web Crawler.
• Keeps Baidu detection scoped to Baiduspider to avoid false positives from unrelated traffic that only contains the company name.
• Replaces flat per-call guesses with a local, privacy-safe ledger built from the token and tool usage returned by each AI provider.
• Adds a monthly spend dashboard with total CiteTrack spend, calls, token usage, search fees, provider breakdowns and daily cost movement.
• Tracks citation checks, AI-assisted fixes, outreach drafts, reputation analysis, brand analysis and prompt discovery without storing prompts or API keys in the cost ledger.
• Uses provider-reported costs when available, dated model pricing otherwise, and clearly marks unknown models as unpriced instead of inventing a number.
• Separates new usage-based totals from pre-upgrade history, which did not contain the token metadata needed for an honest calculation.

• Slack and custom webhook addresses are now encrypted and masked in WordPress.
• Existing webhook settings are automatically moved into the encrypted vault.
• Unsafe webhook destinations are rejected to prevent requests to private or local networks.
• Slack alerts can only use the official secure Slack webhook service.
• Removed insecure reversible storage when server encryption is unavailable. Administrators now receive a clear configuration warning instead.
• Limited MCP requests to 25 operations per batch to protect server resources.
• Rebuilt citation runs around a database-backed queue that remains small, resumable and reliable with thousands of prompts.
• Active runs from the previous queue format are migrated without repeating completed work.
• Citation jobs continue safely after individual failures and stop repeated transport errors from wasting API credits.
• Prompt searching, filtering and pagination now happen directly in the database for faster loading on large prompt libraries.
• Near-duplicate prompt checks are limited to the visible results instead of comparing the entire library.
• Visibility reports now load only the selected reporting period instead of processing unnecessary history.
• Historical visibility data is grouped by day to reduce database and memory usage.
• Crawler response-time calculations now use a bounded recent sample.
• Crawl-waste reporting now uses grouped database queries and avoids running the same analysis twice.
• Attribution and conversion reporting now aggregates data in the database and respects the selected date range.
• “Indexed but not cited” checks now examine only relevant candidate pages.
• Citation receipt reports are bounded by retention settings and sensible row limits.
• Added database indexes for citation runs, won citations, referrals, sessions and conversions.
• Upgraded the database schema to version 16.
• Replaced thousands of individual crawler-verification events with a deduplicated IP and crawler-suffix queue.
• Added limits for pending verification work and rows processed per bucket.
• Excess crawler requests remain safely unverified instead of creating unlimited background jobs.
• Added a database index that speeds up latest citation lookups by prompt and AI service.
• Limited Action Center source analysis to a recent, defensible evidence window.
• Replaced repeated daily chart queries with grouped date-range queries.
• Changed attribution tracking to collect landing-page visits by default, reducing unnecessary requests.
• Developers can still enable all-page or AI-referral-only attribution collection using filters.
• Self-hosted updates must now include a valid SHA-256 checksum.
• Update signatures are required whenever a public verification key is configured.
• Manual citation runs stop after three transport failures to avoid repeated API charges.
• Chart.js now loads only on CiteTrack screens that actually display charts.
• Replaced Gemini’s expensive full-search connection test with a lightweight API-key and model check.
• Real Gemini grounded citation checks can now run for up to 45 seconds.
• Each prompt and AI service remains isolated in its own bounded job.
• Requests that already timed out are no longer immediately retried, preventing a second long delay.
• Gemini timeout messages now explain that the API key may still be valid and identify the hosting-network checks to perform.
• Gemini credentials are now sent securely in the x-goog-api-key header instead of appearing in the request URL.
• Reputation claim analysis now follows the AI Assistant engine and model selected in Settings.
• Removed obsolete hard-coded Gemini 1.5 usage from Reputation and prompt discovery.
• Failed or timed-out AI writing requests no longer apply or add content.
• Failed prompt-generation requests now return a clear error instead of incomplete results.

• Prevented false schema failures when a rendered page cannot be fetched. These checks now show “Couldn’t verify” instead of incorrectly lowering the score.
• Added the exact evidence behind every page check, including detected schema, question headings, external links, metadata lengths and heading counts.
• Added “Check again now” to page tasks. This runs a fresh scan and immediately reports whether the issue passed, still fails or could not be verified.
• Resolved page tasks now complete automatically. They reopen if a future scan confirms that the problem returned.
• Added permanent task notes and a visible activity history for generated drafts, contact checks, verification and follow-ups.
• Improved responsive layouts so long AI service and model names no longer overlap buttons or other controls.
• Added an up-or-down comparison directly to every AI visibility metric.
• Separated brand presence, citation rate and average share of voice into distinct metrics with individual trend charts.
• Trend comparisons now use only prompt and AI-service combinations available at both comparison dates.
• Added notices when changes in tracking coverage could affect a comparison.
• Added trends for average cited-source position, recommendation rank, unique outside sources and cited pages.
• Added gained and lost source-domain analysis without treating every increase as automatically positive.
• Added clear explanations of what each visibility metric means.
• Task generators now follow the AI Assistant engine and model selected in Settings by default.
• Users can still override the AI service for an individual task.
• Added the complete saved evidence directly inside AI visibility tasks.
• Evidence now includes the prompt, AI service, previous and current answers, citation status, cited URLs and cited sources.
• Replaced generic SEO draft options with solutions specific to the detected issue.
• Tasks can now propose metadata, page titles, canonical URLs, internal links, image alt text and page-specific content improvements.
• Added support for editing and applying compatible fixes directly from the task.
• Applied changes remain in WordPress or the active SEO plugin after CiteTrack AI is disabled.
• Internal-link tasks now identify the exact suggested source and destination pages.
• Image tasks now identify the specific images missing alt text.
• A task is created for every failed page check.
• Home remains focused on the three most important AI visibility priorities, while the complete task list is available on the paginated Tasks screen.
• Simplified Home around the three most important AI visibility actions.
• Home now clearly states how many additional visibility tasks are available.
• Removed duplicated Analytics and Crawler reports from Home.
• Added direct links from Home to the complete Analytics and Crawler workspaces.
• Added a dedicated Tasks screen with work-area filters, search, status filters, assignment details and pagination.
• Every action now opens on its own task-detail screen instead of scrolling to a planner at the bottom of a long dashboard.
• SEO, AEO and GEO page improvements now live in Optimize.
• Editorial relationships and outreach work now live in Sources.
• All work continues to share one task and activity history.
• Replaced the unclear “Open all work” path with specific destinations for AI visibility, website improvements, source opportunities and due reviews.
• Added a plain-language Action Center that turns visibility findings, cited sources, claims and page evidence into actionable tasks.
• Tasks can now include an owner, status, expected outcome and review date.
• Added advanced agency filters while keeping the first recommendations simple for local businesses and non-technical users.
• Added AI-assisted creation of evidence-based fix guides, content briefs, correction notes and outreach messages.
• Users can generate assistance with any supported AI service connected to CiteTrack AI.
• Outreach contact discovery now uses only publicly available author profiles, editorial contacts, contact pages or published email addresses.
• CiteTrack AI never guesses an email address.
• AI-generated addresses are removed when they cannot be confirmed in the saved evidence.
• Answer summaries and buyer FAQs are added to the actual WordPress page so they remain after CiteTrack AI is disabled.
• Added support for saving meta descriptions through compatible SEO plugins.
• Schema fixes that depend on another plugin are no longer presented as permanent one-click changes.
• Saved tasks retain their supporting evidence, owner, expected mechanism and follow-up date.
• Completed work can now be reviewed against its original evidence and baseline.

• Recognizes more current AI and search crawlers.
• Verifies supported crawlers using official network information.
• Clearly separates verified crawlers, unverified names and possible impersonators.
• Correctly distinguishes crawler visits from AI access settings.
• Automatically updates the crawler list without storing raw visitor IP addresses.
• Keeps WordPress responsive when outside services are slow.
• Separates human visits from crawler traffic for cleaner analytics.
• Reduces duplicate and suspicious tracking activity.
• Removes private URL details before storing data.
• Runs large scans in the background to reduce timeouts.
• Strengthens permissions around optimization tools.
• Improves data cleanup while preserving saved reports and assigned work.

• Measurement integrity: verified crawler filters, correct queue naming, failed-run coverage, prompt versions, platform denominators.
• Source Intelligence: source URLs, ownership/type, impact, claim association, source shifts and source drill-downs.
• Decision Reports: three decision cards, owners, due dates, evidence appendix and archived methodology snapshot.
• Relationship Workflow: public contact discovery, outreach/correction queue, approval, follow-up date and measured outcome.
• Unified Opportunity Funnel: indexed → verified crawl → citation → referral → conversion, with a separate queue for every broken stage.

Fixed hourly and minute crawler activity display

Fixed LLM models so its more economic for each API calls.

• The crawler chart now has a Day / Hour (48h) / Minute (60m) switch. Hour buckets the last 48 hours, Minute buckets the last 60 minutes (a 90-day minute axis would be meaningless, so those views use their own window and say so), still stacked by engine and still bar/line switchable.
• Grok (xAI) tracking available

1.67.0 ships the Brand Intelligence foundation — all six dimensions, with done-for-you setup. There's a new Visibility → Brand tab that turns the AI answers you already collect into a 360° brand view, using the customer's own AI key (no new data collection).

It opens with a guided 3-step bar so a business owner isn't staring at a blank screen:

1. Brainstorm — you enter your brand (prefilled), your category/niche (so the "best [category]" prompts make sense for an ecommerce store vs a SaaS vs a coach), and competitors (prefilled from what you already track), then "Generate suggestions."

2. Add brand-monitoring prompts (one click) — seeds the questions buyers actually ask AI about a brand: is X legit, X reviews, X alternatives, is X worth it, X pros and cons, what do people say about X, X pricing, and X vs <competitor> (it pulls competitors from your existing lists). This is the "set it up for them" piece.

3. Run them in Citations.

4. Analyze brand mentions — one LLM pass reads the stored answers and extracts the structured signals.

Then it renders all six:

• Presence — recommendation rate + a bar mix of recommended / listed / negative / absent.
• Position — who AI keeps comparing you to, and your average rank when AI ranks options.
• Perception — how AI describes you: positive vs negative attribute chips, plus the audiences AI frames you for.
• Accuracy — the specific claims AI makes about you with accurate/outdated/false verdicts (fed by the Reputation claims audit).
• Provenance — the source domains shaping your AI narrative (you vs third-party). This is the most actionable one: it tells a brand exactly which Reddit thread / G2 page / press article to go win.
• Risk — recurring complaints AI repeats, and engine divergence (ChatGPT vs Gemini disagreeing = unstable narrative).

Added safer duplicate cleanup behavior

Added visible Serve llms.txt setting and save path.

Added optional human request logging, off by default to reduce server load.

Reworked “de-direct” attribution into honest confirmed vs probable attribution with evidence labels.

Added Grok/xAI citation support using grok-4.3 + web search.

Replaced direct Copilot engine claims with Bing Webmaster Tools signals and URL submission.

Added GDPR exporter/eraser hooks and purge-by-date settings action.

Fixed multi-currency revenue display and GA4 conversion currency push.

Added model-aware cost reporting: citation rows now store model, and Costs groups by engine/model.

Updated current defaults: gpt-5.4-mini, gemini-3.5-flash, claude-sonnet-4-6, sonar-pro, grok-4.3.

CiteTrack AI updater now refreshes activation tokens, avoids 6-hour caching for bad update responses, clears WordPress update cache when license state changes, and adds a Check for update action on the Plugins page.

For GSC, I added a real property dropdown (it was a disabled, auto-only list before).

For GA4, the override dropdown already exists and lists every property with its ID

Overwrite-in-place. The updater now detects when a package is ours (your uploaded zip, checked by looking inside it) and forces WordPress to overwrite the canonical citetrackai folder rather than spawn a duplicate. So future uploads of any citetrackai-X.Y.Z.zip replace in place.

One-click cleanup. The duplicate warning on the Plugins screen now has a "Clean up duplicate copies now" button that deactivates and deletes every other CiteTrack AI copy, keeping the running one.

Richer dashboard — rebuilt Overview: real KPI tiles, a 30-day AI-sessions trend chart, a by-platform breakdown, and top AI landing pages ranked by revenue.

Per-page journeys — a new Journeys tab follows individual AI sessions (landing → every page viewed → conversion), backed by a new referral_hits table.

Competitor benchmarking — "Who AI cites in your space," ranking every domain (you highlighted) by citation share across your tracked prompts.

Custom goals & values — a Goals tab to define URL-visit or coded-event goals with a dollar value; AI-referred sessions that hit them count as conversions (do_action('citetrack_goal','slug') for code-fired events).

• • Smarter scoring by content type. Optimize now scores each post type with the right profile — articles get the full AEO/GEO checks, while pages (landing/legal) and products (WooCommerce/EDD) only get checks that actually apply, so irrelevant items no longer drag scores down. Also: the dashboard now explains exactly why GA4 may still show server-log data instead of real sessions (Google not connected / no property picked / property not readable).
  • Duplicate-install protection. If a second copy of the plugin ever gets installed (e.g. a release zip extracted into its own folder), it now loads as a harmless no-op, deactivates itself, and tells you which folder to delete — instead of fataling with "Cannot redeclare class."
  • Fixed "Connect with Google" 403. The licensing-server hardening had started rejecting the Google OAuth relay calls; the plugin now sends valid license credentials on connect, claim, and the hourly token refresh. This also fixes "connected but no Search Console properties / no GA4 data."
  • UI parity (design mockups → product). License screen now shows sites used, renewal date, and a "what your license unlocks" checklist; Alerts gets a cleaner template dropdown plus delivery channels (email, Slack, signed webhook); Settings rebuilt as a modules + data/privacy layout; Reports moved to a two-column layout.
  • Hotfix. Properly stopped the "Cannot redeclare function citetrack()" fatal by guarding the function at compile time (the 1.56 guard ran too late). Plus clearer guidance to always upload the shipped zip, never a re-zipped folder.
  • One-click internal links. Internal-link suggestions now have an "Insert link" button that writes a real link into the post content (first plain-text match, skipping existing tags/links), with a reversible undo stack.
  • Cross-type internal links. Those suggestions now span every public post type — a post can suggest linking to pages, products, or custom post types, each labeled by type — not just other posts.

• • The analyzer now fetches the rendered permalink HTML—allowing it to detect third-party JSON-LD and fall back to the rendered head for meta descriptions—while utilizing a 12-hour cache and strict request budgets to prevent server hangs.
• • The "Top fixes" column now displays actionable instructions instead of confusing pass-condition labels, and introduces an "all checks pass" pill when there is nothing left to fix.
  • The score badge for values of 300% or higher is now styled in solid brand indigo instead of amber.
  • The cost ledger and per-call rate estimates have been moved from the Insights tab to a dedicated Settings → Costs screen, which includes built-in redirection notices for legacy links to prevent dead ends.

• Integrated GA4 for accurate human traffic tracking,
• Upgraded the dashboard with dynamic filtering and new health charts,
• and introduced a Page Value table with automated insights comparing AI versus human engagement.