How to Audit Your Site’s AI Visibility in 2026
Quick answer: A complete, repeatable AI visibility audit. The three layers to check (access, readiness, performance), the exact signals that matter, and how to run and package it with CiteTrack AI.
⚡ Key takeaways
- → An AI visibility audit is a structured one-time measurement of how AI engines treat your site. Five phases: prompt research, baseline citation measurement, crawler reach, source pattern analysis, action plan. Run it once, then move to ongoing monitoring.
- → Most websites are nowhere near ready. Foglift's Q2 2026 AEO Readiness study of 1,386 scans across 344 domains found the median AI Readiness Score sits at 46 out of 100 versus a median SEO score of 86 out of 100. Nearly half of SEO-strong domains still scored below 50 on AEO.
- → The slots are shrinking. Resoneo's study of 27,000 ChatGPT responses found the March 2026 model switch dropped unique domains per answer from 19.1 to 15.2, a 20.5% contraction.
- → Citation behavior is volatile. AirOps 2026 found only 30% of brands stay visible from one AI answer to the next. One-time snapshots tell you almost nothing.
- → The audit reveals what to fix. Ongoing monitoring tells you whether the fixes worked. Different disciplines, different cadences.
Why an audit beats another SEO checklist
You probably know your organic rankings cold. You watch keyword positions weekly. You review backlink profiles. You have GSC bookmarked.None of that tells you whether ChatGPT cites your site, whether Perplexity links to your pricing page, or whether Gemini recommends you to buyers.
An AI visibility audit answers a different question: when an answer engine builds a response to the question your buyer actually asks, where does your site fit?
The readiness gap (Foglift, 344 domains, Q2 2026)
Nearly half of SEO-strong domains scored below 50 on AI readiness. The gap is structural.
The unit of analysis is the prompt, not the keyword. "Best CRM for small teams" replaces "best CRM small teams." Buyer intent gets expressed in full sentences now, and AI engines weight that structure differently than Google did.
The metric is presence in answers, not position in results. Either your brand is named in the response or it isn't. There is no #8 in an AI answer; there are roughly three to five brands cited, and the rest of the category is invisible.
The sources AI engines cite are different from the URLs that rank organically. Third-party reviews, community threads, niche directories pull more weight than they did in classic search. The audit measures that explicitly.
Everything below is the framework. Five phases, run once as an audit, then converted to monthly cadence afterward.
The five-phase audit
What buyers ask
Cited, mentioned, absent
Who reaches you
Where trust lives
Prioritized fixes
Phase 1: Prompt research
Your audit starts by figuring out which prompts your buyers actually run. Not the keywords your team thinks matter. The questions buyers type into ChatGPT, Perplexity, Gemini, and Claude.
Start with 50 prompts. Twenty to thirty is the floor for a directional read; 50 to 100 gives statistical reliability. Mix three intent types.
Category awareness: "What is AI visibility tracking" / "What is an answer engine"
Solution comparison: "Best AI citation tracker for small teams" / "Profound vs AthenaHQ"
Decision intent: "How much does Profound cost" / "Is AthenaHQ worth it for a 5-person team"
Run them across the engines your buyers actually use. ChatGPT, Perplexity, Google AI Mode, Google AI Overviews at minimum. Add Claude and Gemini if your audience trends technical or enterprise.
Document for each prompt: which brands appeared, in what position within the answer, and which sources the engine cited.
Our 101-question benchmark across 4 engines shows what normal citation rates and source counts look like, so you have a baseline before you start measuring your own.
Phase 2: Baseline citation measurement
Now you measure how often your site is cited across the prompts you researched.
Run each prompt three to five times. Citation behavior is non-deterministic. AI engines sample sources from a probability distribution; the same prompt produces different answers on different runs. One snapshot is noise. Three to five runs give you signal.
Calculate three metrics for each prompt.
Mention rate: how often your brand is named in the response, regardless of citation versus mention.
Citation rate: how often a URL from your domain appears as a cited source.
Position: where in the answer your brand appears (first sentence, middle, last).
Expect volatility. AirOps's research, cited above, found only 30% of brands stay visible answer-to-answer and just 20% across five consecutive runs. If your brand is cited in two of five runs, that's not stable visibility; it's a high-variance presence. The baseline measurement should reflect that variance, not hide it.
The output of Phase 2 is your starting line: average citation rate across the prompt set, breakdown by engine, position distribution.
If you want to understand why certain pages keep getting skipped, your Citability score breaks down the structural signals engines look for when picking sources.
Phase 3: Crawler reach audit
Even with great content, if AI crawlers can't read your site, none of it matters.
Check three things.
Which AI bots actually visit your site. GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, and others should appear in your server logs. If they don't, your crawl signal is weak and your content isn't reaching the training and retrieval pipelines.
Whether your robots.txt allows them. Many sites accidentally block AI crawlers via blanket User-agent rules combined with strict disallow patterns. Worth checking line by line.
Whether your content renders for non-browser crawlers. Resoneo's crawler rendering study found that ChatGPT, Claude, and Gemini are all HTML-only crawlers with no JavaScript execution. Pages that depend on client-side JS to display content appear empty to them. Only Copilot executes JavaScript.
GA4 will not surface any of this. Google Analytics is client-side, it relies on JavaScript execution to record visits, and AI bots do not run JavaScript. The entire bot category falls outside GA4's visibility. The audit needs server-log access. OtterlyAI's analysis of 1 million+ citations found 73% of sites have technical barriers blocking AI crawler access, so this phase catches problems most sites do not know they have.
On WordPress, the crawler tracking walkthrough has the step-by-step setup with reverse-DNS verification.
The result is a list: which engines reach your site, how often, and which pages they actually visit.
Phase 4: Source pattern analysis
You take the citations your audit surfaced and analyze which types of sources AI engines treat as authoritative for your category.
Build a source pattern map from your actual citation data.
Owned content: your site, your blog, your documentation. What percentage of citations point here? In most B2B categories, the answer is well under half. Your owned content matters, but it does not carry the weight it used to.
Review and comparison platforms: G2, Capterra, TrustRadius, Trustpilot, Software Advice. These dominate in B2B SaaS buyer prompts. If your category has heavyweight review sites, expect them to appear repeatedly.
Community platforms: Reddit, Stack Overflow, niche subreddits, forums. OtterlyAI's million-citation study found community platforms capture 52.5% of all citations versus 47.5% for brand domains. That number alone explains why on-site content is only half the battle.
Editorial coverage: industry publications, trusted blogs, trade press. Authority compounds here.
Aggregators and directories: industry-specific listings, "best of" roundups, software directories.
The Princeton GEO study (Aggarwal et al., ACM KDD 2024) tested nine optimization techniques against a 10,000-query benchmark. Their three top-performing methods, adding statistics, adding citations, and adding direct quotations, achieved a relative 30 to 40% improvement on the Position-Adjusted Word Count metric. The implication for your audit: pages that include verifiable data and explicit source attribution are more citation-eligible than pages that don't.
Freshness matters too. AirOps's report found more than 70% of pages cited by AI were updated within the last 12 months, and 83% of commercial citations come from pages refreshed within the past year. A 36-month-old pricing page is a citation liability even if the price is still accurate.
Where AI citations actually come from (OtterlyAI, 1M+ citations)
More than half of all citations go to sources you do not control. If your audit shows you depend heavily on owned content while competitors get cited across reviews, communities, and editorial, that is the diagnosis.
Phase 5: Action plan
Your audit produces three lists. The action plan converts them into work with timelines.
Easy wins (week 1 to 4): pages where you're already close. Citation present on 2 of 5 runs, brand named but not linked, pages indexed but missing structured data. Small fixes move these into consistent visibility.
Structural fixes (month 1 to 3): pages where the content is good but the format isn't extractable. Add FAQs, restructure headings to mirror buyer questions, add tables for comparison content, add author bylines with credentials.
Authority gaps (month 3 to 12): categories where you're invisible because the third-party signal isn't there. Pursue review profiles, contribute to community threads, pitch comparison roundups, secure analyst inclusion. These are the slowest fixes and the highest-leverage ones.
For prioritization, score each item on impact and effort. Move the high-impact / low-effort items first; defer the high-impact / high-effort ones to a quarterly planning conversation.
The audit ends here. Ongoing monitoring is a separate discipline. The audit gives you the picture; monitoring keeps you honest about whether the work moved the metric.
Once the audit is done, ongoing competitor monitoring keeps you honest about whether the work moved the metric.
What the audit will not tell you
Three things the audit deliberately won't answer.
Whether AI traffic is driving revenue. That is a separate analysis. If you sell through WooCommerce, the AI sales tracking setup handles the attribution side.
Why AI engines are not citing you specifically. The audit tells you that you are not cited. The diagnostic for common failure modes is the next step for root-cause analysis.
Which AI visibility tool to buy. The audit runs on free tools and manual prompt runs. If you decide to invest in tooling afterward, the citation tracker comparison has current pricing and tradeoffs at every budget.
Frequently asked questions
How long does an AI visibility audit take?
Plan for 17 to 30 hours of work, depending on prompt-set size and engine coverage. Phase 1 (prompt research) is the most variable. With a focused prompt set, the audit completes in roughly two working weeks of part-time effort.
How often should I re-audit?
Annually for a full audit. Quarterly for a lighter version that re-runs your priority prompts and updates your source pattern map. Ongoing monitoring lives in between, weekly or biweekly.
What is the minimum viable audit?
20 prompts, three engines (ChatGPT, Perplexity, Google AI Overviews), three runs per prompt. That is enough to surface the major gaps. Scale up to 50+ prompts and 5 engines when you have the budget for the rigor.
Can I run the audit on free tools?
Yes. Manual prompt runs against ChatGPT, Perplexity, and Gemini cost zero and tracking is in a spreadsheet. Paid tools speed up the runs and add competitor comparison, but the framework is the same.
What is the difference between an audit and an analytics review?
Different data sets. The audit measures presence in AI answers, which is zero-click visibility. An analytics review measures clicks that landed on your site. AI search produces results without clicks, so the audit catches the visibility you would miss in GA4.
Ready to run your audit? CiteTrack gives you citation tracking, crawler logging, and source-pattern analysis to run the audit and ongoing monitoring on one license. Solo $59/year for a single site, Studio $129/year for up to 5 sites, Agency $249/year for unlimited sites.
See it in your own WordPress dashboard
Find out if AI recommends you — or a competitor
Install CiteTrack AI and see exactly who crawls you, who cites you, and what it earns. Flat pricing from $59/year, 14-day money-back guarantee.
Want to see if AI cites pages like this?
CiteTrack AI tracks your AI crawlers, citations, and the revenue they drive, inside WordPress.
Explore: Features · Crawl Academy · Glossary · Compare · Pricing