We rank on page one for most of our target keywords, but ChatGPT and Perplexity cite our competitors instead of us. Why?
Ranking on page one but not cited by ChatGPT or Perplexity? Here is the diagnostic order we use with clients, and the Promptwatch features we rely on for each step.
This is one of the most common problems we see when a client first comes to us for AI visibility. Strong Google rankings, solid traffic, and then a screenshot from a founder: ChatGPT recommended three competitors and not them. The reaction is usually that something is broken. Usually nothing is broken. The AI path simply uses different signals than the Google path, and the gap is diagnosable.
Here is the order we work through, and why we run it in that order.
Step 1: confirm the pattern with a proper sample
One screenshot proves nothing. We start by tracking the client's target prompts across ChatGPT and Perplexity for a couple of weeks, so we can see the rate at which competitors are cited instead of the client. Answers vary run to run, and a single miss may not be a pattern.
We use prompt panels in Promptwatch for this, split by prompt type and persona, so we can see whether the gap is on comparison questions, problem-led questions, or brand questions. The shape of the gap often hints at the cause.
Step 2: check whether the AI crawlers can reach the site
This is the first technical check, because it is quick and the fix can be fast. Google can crawl a site perfectly while GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, or PerplexityBot are blocked by a robots.txt rule, a CDN setting, or bot protection that treats unknown agents as threats.
We look at Agent Analytics, which shows IP-verified AI crawler logs and filters out spoofed user agents, with error tracking and failure explanations. A page that ranks first on Google and returns a 403 to an AI bot will never be cited, and no amount of rewriting helps.
Crawler logs start on the Professional plan, so for clients on Essential we either upgrade for the audit period or check the logs by hand through their CDN.
Step 3: look at the fan-outs
AI engines rarely run the client's target keyword as typed. They break a question into sub-queries and retrieve results for each. The client may rank well for the head term and appear in none of the fan-outs.
We pull query fan-outs per model for the prompts where the client is missing, then check where the client ranks for those longer phrasings. Often the answer is "nowhere," because the site has a page for the keyword and nothing for the specific question inside it.
Step 4: see who is actually being cited, and from where
We then look at citation analytics at page and domain level. In many categories the cited pages are not the competitors' own sites. They are review platforms, community threads, comparison articles, and listicles. If competitors show up across those and the client does not, then the competitor's visibility is built off-site.
Offsite mentions (a brand named in a third-party page without a link) are especially useful here, since they catch references that backlink tools miss. Reddit, YouTube, and LinkedIn citation reports show the community side.
This is also the step where expectations need managing. Earning presence on third-party pages is slower than rewriting a page. We tell clients so, and we plan it in parallel with the quick fixes.
Step 5: compare the page shape
If the crawlers can reach the page and the fan-outs are covered, we compare the client's page with the cited competitor page for the same question. We look for a direct answer near the top, clear structure, a table or numbered process where the question invites one, and specifics instead of general copy.
Content gap analysis with coverage scores helps us see which questions the client's site does not answer at all. For existing pages, Content Refresh produces diff-style suggestions, which is easier to review than a full rewrite.
Step 6: join it all per page
To see ranking and citation side by side, we use 360 Insights, which joins crawler logs, citation clicks, and AI-referred visits per page, and Search Insights, which overlays Google Search Console and Bing Webmaster traffic on AI crawls and citations. For a page that ranks on Google but is not cited, that view shows whether the problem is access, content, or off-site presence.
What we do with the findings
The Action Board converts the diagnosis into a prioritized list with owners. For content work, the platform can draft pages against the gaps and publish to WordPress, Webflow, or Framer after review. The automated Content Agent is on the Business plan, so for smaller plans we use on-demand content creation instead.
What we tell clients to expect
- Crawler fixes can change results quickly, and they are the first thing to try.
- Content changes take longer, because engines must re-fetch and re-evaluate.
- Off-site presence takes the longest and cannot be bought by the dozen.
- No tool or agency can guarantee a citation. Engines choose their own sources.
If you rank on page one and are not cited, the answer is almost always in one of the places above. Measure first, check crawler access second, then fan-outs, sources, and page shape. That order saves the most time, and it is why our client work for this problem runs on Promptwatch.