Why Isn’t My Site Showing Up in AI Answers in 2026

You set up your buyer prompts. You checked ChatGPT and Perplexity. Your URL is not in the citations. Something is wrong. The hard part is knowing what.

This is a diagnostic post. It walks through the seven causes that explain almost every "I am not getting cited" case, in the order you should check them, with the test and fix for each.First, some context on what normal looks like. In our 101-question benchmark, AI engines cited a source 95% to 100% of the time. But the number of slots per answer varies wildly.

There are slots. The question is why you are not in them.

Where to start the diagnostic

Work top-down through the causes. Most teams jump to "improve content quality" first because it feels productive. That is usually the wrong starting point. If AI crawlers cannot reach your pages, no amount of content quality work moves the needle.

The right order. Start with reach (can bots see your pages?), then match (do your pages answer the prompt?), then structure (is the answer extractable?), then authority (do engines trust your domain?). Each layer depends on the one below it. Skipping levels wastes weeks.

What data do you need to run this? Three things. Whether AI bots crawl your pages (server logs or a WordPress plugin). Whether you are cited on your buyer prompts (a citation tracking tool). What your citability score looks like per page (a scoring tool, manual audit, or both). The first two are essential. The third is helpful.

If you do not have any of this data yet, set up tracking before the diagnostic.

. The guide to setting it up is in how to track AI citations.

Cause 1: AI crawlers cannot reach your pages

The most common cause. AI engines cannot cite pages they have not read.

Why does this happen? Several reasons, and any one of them is enough.

• Robots.txt blocks AI bots. Many sites still have Disallow rules from the early days of AI scraping concerns. GPTBot, ClaudeBot, and PerplexityBot all respect robots.txt. If your file blocks them, they do not crawl.
• JavaScript-rendered content. AI crawlers fetch HTML directly. They do not execute JavaScript. If your page content loads after the initial HTML response (via React, Vue, or similar), AI engines see an empty shell.
• No llms.txt declaration. Not a hard block but a signal. Sites without llms.txt are harder for engines to prioritize.
• Server response issues. 503 errors, slow response times, or aggressive rate limiting can cause bots to give up.
• Cloudflare or CDN bot challenges. Some sites have bot protection that misidentifies AI crawlers as malicious and blocks them.

How to test. Read your server access logs and look for GPTBot, ClaudeBot, PerplexityBot, Google-Extended user-agents. If you do not see them at all, bots are not reaching your site. If you see them but only on a few pages, the rest of your site is invisible to AI.

How to fix.

• Edit robots.txt to allow AI bots. Specifically allow GPTBot, ClaudeBot, ChatGPT-User, OAI-SearchBot, PerplexityBot, Google-Extended, and Anthropic-ai. Do not Disallow them.
• Ensure server-side rendering for content-critical pages. If your CMS renders content via JavaScript, switch to SSR or static generation for pages you want AI to read.
• Generate llms.txt and llms-full.txt at your site root. CiteTrack and some SEO plugins handle this automatically on WordPress.
• Check your CDN bot protection settings. Whitelist verified AI crawler IPs if your CDN blocks them.
• Submit your sitemap to engines that accept submissions (Bing for Copilot, Google for AI Overviews).

User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

User-agent: PerplexityBot
Disallow: /

User-agent: GPTBot
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: PerplexityBot
Allow: /

One stray Disallow line is the most common reason a site goes invisible to AI. Check yours right now: open yoursite.com/robots.txt in a browser.

Cause 2: Pages reach but content does not match prompts

AI crawlers are reaching your pages. Engines have the content. But you still do not get cited.

Why? Usually because your content does not match the buyer prompts your tracking checks. The intent is wrong.

How to test. Pick five buyer prompts from your tracking set. Read them carefully. Now read the page you expect to get cited on each. Ask yourself: does this page actually answer this prompt?

The common mismatches.

• Page is about your product, prompt is about the problem. "Best CRM for small business" is a problem-led prompt. A page titled "Why XYZ CRM is the right choice" is product-led. The mismatch is structural.
• Page covers a tangent, not the core question. Prompt asks "how much does X cost." Page discusses cost considerations broadly without giving an actual price range.
• Page is competitive content for a head term. Prompt is specific ("CRM for B2B SaaS with under 50 employees"). Page is generic ("CRM software"). Generic pages lose to specific pages.
• Page exists but engines cite a competitor's version of the same content. The prompt has a "right answer" that lives somewhere else. You need to either match that answer or differentiate clearly.

How to fix.

• Rewrite the opening paragraph to answer the prompt directly. If the prompt is "best CRM for B2B SaaS," the page should open with "The best CRM for B2B SaaS is X, because Y."
• Audit the H2 hierarchy. Each H2 should answer a sub-question that a buyer would naturally ask. Random H2s waste extraction surface.
• Match the prompt's specificity. If buyers ask about "under 50 employees" specifically, the page should address that scope, not just generic "small business."
• Create new content for prompts you do not currently cover. If 5 of your 20 prompts have no matching page, you cannot expect citations on those 5.

Cause 3: Match but structure is hard to parse

The content matches the prompt. The intent is right. But you are still not getting cited. The structure is the problem.

AI engines extract answers from clean structures. Dense paragraphs, missing headings, no schema, no clear definitions, no direct answer up front. All of these make extraction harder. Engines pick easier sources.

How to test. Read your page as if you were a parser. Can you extract a one-paragraph answer to the buyer prompt without reading the whole post? If yes, structure is fine. If no, the structure is the problem.

Specific structural issues to check.

• The direct answer is buried. The first paragraph sets context, talks about why the topic matters, gives background. The actual answer appears in paragraph 5 or 6. AI engines see paragraph 1 and use it. The actual answer is invisible.
• Headings do not summarize sections. "Things to consider" or "Important points" tell engines nothing. Headings should be like "Best CRM for B2B SaaS in 2026" or "Why X costs $99/mo," extractable on their own.
• No schema markup. Engines use FAQPage, HowTo, Article, Product schema as parsing hints. Pages without relevant schema lose to pages with it.
• Walls of text. 600-word paragraphs without breaks. Engines parse shorter, structured chunks more easily.
• Lists used inconsistently. Some sections use bullets, some use prose, no pattern. Inconsistent structure confuses parsing.

How to fix. Open the page. Move the direct answer to the first paragraph. Add H2 headings that summarize each section. Add FAQPage or HowTo schema if applicable. Break long paragraphs. Use bullets for unordered information, numbered lists for ordered steps. Most pages can be structurally fixed in 30 to 60 minutes. Your Citability score shows exactly which signals each page is missing.

Cause 4: You are cited but on prompts you do not track

This one is sneaky. Your tracking shows no citations on your tracked prompts. So you assume you are not being cited at all. But AI engines cite you on prompts you are not measuring.

How to test. Manually ask AI engines a wider set of questions in your category, including ones not in your tracking set. Look for your URL in the citations. If you see citations on questions outside your tracking set, the issue is measurement coverage, not citation absence.

Why this happens. You picked 15-30 prompts for your tracking set. AI engines are answering thousands of related questions every day. Some of them might be citing you on prompts you did not think to track. Your dashboard does not show those because your dashboard does not query them.

How to fix.

• Expand your prompt set. Add variations, long-tail questions, comparison prompts, troubleshooting queries.
• Audit referrer-less traffic for evidence of citations. If a specific deep page suddenly gets direct traffic on a topic, that is usually an AI citation kicking in. Reverse-engineer which prompt might have triggered it.
• Use a tool with broader prompt coverage. Some tools (Profound, Peec, Trakkr) support hundreds or thousands of prompts per plan. Manual or smaller sets miss long-tail citations.

Cause 5: Competitors win the category at scale

You are doing everything right. Pages crawl. Content matches. Structure is clean. You are still not getting cited because someone else always gets cited first. Sometimes it is not just about being outranked — AI may be describing your brand inaccurately or framing you for the wrong audience entirely. A brand intelligence check shows what AI actually claims about you and where that narrative comes from.

This happens in mature, competitive categories where a few sources dominate. CRM, project management, SEO tools, web hosting, email marketing. These have entrenched leaders that AI engines cite by default.

How to test. Look at the citations on your top buyer prompts. If the same 3-5 sources appear in every answer, the category is saturated. Your job is harder than the previous four causes suggest.

How to fix.

• Find sub-categories where dominance is weaker. "Best CRM" is saturated. "Best CRM for solo consultants" or "best CRM with free WhatsApp integration" might not be. Long-tail beats head terms in saturated categories.
• Get cited by the citers. If G2 and Reddit dominate citations in your category, focus on G2 reviews and Reddit presence as primary distribution rather than your own site. AI engines often cite Reddit threads and review platforms heavily.
• Build content the leaders do not have. The dominant sources usually have generic content. Specific case studies, benchmark data, original research, and detailed comparison pages outflank generic pages even on saturated terms.
• Be patient. Even in saturated categories, citation share shifts over months as engines refresh their corpus. Sustained quality work eventually moves the needle.

Cause 6: Domain is too new or low-authority

AI engines weight domain authority signals. A page on a 6-month-old domain with no backlinks competes against pages on 10-year-old domains with thousands of backlinks. The new domain loses, even if the content is better.

How to test. Check your domain age, total referring domains, and overall authority via any SEO tool (Ahrefs, Semrush, Moz). Compare against the domains that get cited on your prompts. If the gap is significant, authority is the cause.

How to fix. Slowly.

• Build authoritative backlinks. Real editorial backlinks from authoritative sites in your category. Paid link networks do not work and may hurt.
• Publish primary research and data. Original benchmarks, case studies, and survey data attract natural backlinks and citations.
• Be cited elsewhere first. AI engines cite sources that other sources cite. Getting referenced in industry publications, podcasts, and authoritative blogs builds the cross-citation graph that AI engines weight.
• Time helps. Domain age is a real signal. A 3-year-old domain on the same content beats a 6-month-old domain. There is no shortcut.

Setting expectations. New domains usually take 6-18 months of consistent quality publishing before AI engines start citing them meaningfully on competitive prompts. That timeline is realistic.

Cause 7: Your topic is in a no-cite zone

Some topics AI engines deliberately under-cite or refuse to recommend on. Medical, legal, financial, and political topics often get hedged responses that cite official sources only.

How to test. Ask your buyer prompts and look at the answer. If engines hedge ("consult a professional," "talk to your doctor," "speak with an attorney") and only cite government or major institutional sources, your topic is in a no-cite zone.

What to do. Honestly, less than you might hope.

• Reframe the content angle. Engines that will not cite you on "best treatment for X condition" might cite you on "questions to ask your doctor about X" or "what to expect at the X appointment." Practical, non-prescriptive content survives the filter.
• Build E-E-A-T credentials hard. Author bylines from licensed professionals, citations to peer-reviewed sources, transparent editorial standards. The "Your Money or Your Life" signal Google uses is similar to what AI engines look for in these zones.
• Accept that some prompts will not move. If the prompt is "what medication should I take for X," no AI engine will recommend your content over a doctor or institution. Pick prompts you can win.

The full diagnostic workflow

Here is the whole thing as a checklist. The order matters. Do not skip levels.

1. Step 1. Check whether AI bots are reaching your pages. Read server logs, install a plugin that hooks WordPress requests, or check CDN bot analytics. If bots are not reaching you, fix that first. Stop everything else.
2. Step 2. Verify content-prompt match on your top 10 prompts. Read each prompt, read each target page. Confirm the page actually answers the question. Rewrite where it does not.
3. Step 3. Audit structure on the same 10 pages. Direct answer up front, H2 hierarchy, schema markup, scannable formatting. Fix where weak.
4. Step 4. Expand your tracking prompt set by 2-3x. Look for citations you may be earning on prompts you do not currently watch.
5. Step 5. Audit the competitive landscape. If your category is saturated, plan a sub-category or long-tail strategy. If a few leaders dominate, identify how to be cited by the citers (G2, Reddit, industry publications).
6. Step 6. Look at authority signals. Domain age, backlink profile, brand mentions. Plan slow-build authority work.
7. Step 7. Consider whether your topic is in a no-cite zone and adjust expectations accordingly.

The work is sequential. Step 1 has to be solved before Step 2 matters. Step 2 before Step 3. Skipping levels is the most common reason teams burn months on the wrong fix.

.

FAQ

Where to go from here

Once you know which cause is yours, the fix usually lives in one of these. If the problem is structure, start with your Citability score to see exactly what each page is missing. If the problem is reach, set up server-side crawler tracking so you can see which bots actually show up. If you need the full playbook, the AEO guide and GEO guide cover on-site, off-site, and measurement end to end. And if you want to track citations across all three layers in one place, that walkthrough has the setup.

Guides · AEO, AI Search, How-To, Troubleshooting