We rank first on Google. Why does ChatGPT not mention us?

Your marketing manager has a slide that says you rank first for the terms that matter. It took three years and a good deal of money to get there, and it still brings enquiries. Then a prospect mentions, almost in passing, that they asked ChatGPT for a shortlist and your firm was not on it. A competitor you regard as smaller was.

ID-Pal, the identity verification firm, found itself in a version of this in May 2026. Across the questions its buyers actually ask, the assistants knew who it was and still passed it over: "AI could describe ID-Pal. It never recommended it." After pages were rebuilt to answer those questions, the count moved: "UK buyer questions where AI names ID-Pal: none in May, two in August, four in September 2026."

Google's results page ranks pages; AI assistants answer questions and name firms. Ranking helps, because most assistants start from a search index, but it is no guarantee. Ahrefs found in March 2026 that only 38% of AI Overview citations came from pages ranking in the top ten, down from 76% in July 2025, because Google now runs several related searches for one question and cites pages from all of them.

#Why does ChatGPT recommend my competitor when we rank higher on Google?

Because a ranking is a position in a list of links, and a recommendation is a name in a sentence. The assistant uses search results to find pages, then decides for itself which firms it can confidently name, and a clearer page from a lower-ranked firm can win that decision.

Ask Google for "employment law solicitors Manchester" and you get a list to choose from. Ask ChatGPT "which firms should I talk to about an employment tribunal in Manchester" and it searches, reads what it fetches alongside what it already knows, and names three or four firms with reasons. The assistant has already chosen.

So a firm can hold the top position and not be named, while a firm at position seven with a clearer site and a couple of trade press profiles is named instead.

#What are AI assistants actually reading about you?

They read your website, the sites that mention you, the reviews and directories that describe you, and anything your named people have published, and they look for agreement.

They have two sources: training data, a snapshot of the web taken at some point in the past, and live retrieval, where ChatGPT search, Copilot, Perplexity, Gemini and Google's AI features fetch current pages using crawlers that behave differently from Google's (see how AI crawlers read your website).

Links still count at the retrieval stage. Live retrieval starts from a search index, and links help decide which pages that index surfaces for the assistant to read. An SE Ranking study reported in November 2025 found the number of referring domains was the strongest predictor of whether ChatGPT cited a page, and AirOps found in April 2026 that pages in the top position of the search results ChatGPT retrieved were cited 58.4% of the time, against 14.2% for pages in position ten.

Once it has the page, the assistant is trying to establish a few facts:

• What does this firm do, in plain words, and for what kind of client?
• Where does it operate and at what scale?
• Who are the people, and what are they known for?
• Do other sources agree, or does the picture contradict itself?
• Is the page an honest answer to the question, or a page built to catch a keyword?

A page that ranks but fails those tests gets read and not named. There is more on the underlying judgement in what E-E-A-T is and why AI cares about it.

#Does ranking on Bing matter for ChatGPT?

Yes, more than most firms assume. OpenAI says ChatGPT search uses third-party search providers, Bing has been reported as one of them, and a small Seer Interactive test in February 2025 found 87% of ChatGPT search citations matched Bing's top results. Microsoft Copilot, which many buyers have at work, runs on Bing.

A firm that ranks well on Google but is thin or missing on Bing may never be fetched by ChatGPT search or Copilot.

Three checks cover most of it. Verify the site in Bing Webmaster Tools, submit your sitemap and see which key pages Bing has indexed. Search Bing for your three most important buyer questions. Then check that robots.txt and your CDN let OAI-SearchBot and Bingbot in: OpenAI says a site must allow OAI-SearchBot to appear in ChatGPT search, and Cloudflare changed its default to block AI crawlers in July 2025, so check the setting.

#Why do keyword-built pages read badly to an assistant?

Because the things that win a ranking and the things that make a clear answer are different, so the pages built hardest for position are often the least likely to be cited.

For example, take a typical ranking page: 2,500 words on "cloud backup for accountancy firms", the phrase in every heading, an introduction that circles the topic for three paragraphs, a generic list of benefits and a call to action every few hundred words. Years of link building keep it at the top.

Now read it as an assistant would, trying to answer "who provides cloud backup for a twelve-partner accountancy practice in the North West". The page does not say who the firm serves or at what size. It does not say what makes its approach different. It does not name a person. It reads as marketing. The assistant moves on to a competitor's page that says, in plain sentences, "we provide backup and disaster recovery for accountancy practices with five to fifty staff, mostly in the North West, and here is how a typical engagement runs".

The second page might never rank first. It answers the question, so it gets named. The same pattern explains why ChatGPT never quotes most blogs.

To see this for your own firm, record the same things for each buyer question:

#Do AI Overviews cite pages outside the top ten results?

Yes, most of the time. In Ahrefs' March 2026 data, 38% of AI Overview citations came from top-ten pages, 31.2% from positions 11 to 100 and 31.0% from beyond the top 100.

The reason is what Google calls query fan-out. Google explains that its AI features generate a set of related searches to answer one question, then draw on pages from all of them. A page that answers a follow-up question well can be cited even if it ranks nowhere for the main search. That is why a competitor who has written plain answers to the questions buyers ask next can appear in the AI Overview above your first-place link. AI Mode goes further, as we explain in what Google AI Mode is and how it differs.

#What are the signs that AI is already costing you enquiries?

Look for these in the monthly report, and add "an AI assistant" as an option on your enquiry form:

• Rankings are stable but visits are falling.
• Click-through rates for the same positions are lower than a year ago.
• Enquiries from search are fewer, or weaker, without an obvious cause.
• Prospects say they "looked you up" in ways your analytics cannot explain.

If two or more are true, the ranking is already worth less than the slide suggests. We cover the results page itself in why traffic falls when rankings have not.

#What actually protects you?

Being the clearest, best evidenced answer to the questions your buyers ask.

#Expertise that is specific to you

A method, a point of view, original research or a very specific description of who you serve is hard to substitute. Be narrow enough that only you could have written it, and put a named person's name to it: assistants attach expertise to people as well as firms, and "the team" is not someone they can recognise.

#Evidence from other people

Trade press profiles, directories, reviews, awards and case studies with named clients tell an assistant that someone other than you says you are what you claim. Consistency matters more than volume: the same facts, in the same words, across several sources.

#A website that answers questions

Every service page should say, in its first hundred words, what you do, for whom, where and at what scale, with the key text in the page's HTML so crawlers that do not run JavaScript can read it. ID-Pal's work followed this pattern: "Five industry pages rebuilt: fund administration, wealth management, legal, accountancy and estate agency. Each opens with the question a compliance lead types, names the regulator, and answers inside the first ninety words." The result, in the case study's words: "Where a page has been rebuilt, ID-Pal appears. Where it has not yet, it does not, which tells us exactly what to build next."

#Channels you own

A referral network and a direct relationship with your market do not depend on how strangers find you. Researched outbound, where a person writes and approves every message, reaches buyers before any assistant is asked.

#What should we do with the rankings we already have?

Use them to guide the shift. Your rankings tell you which questions your market asks and which of your pages the search engines already trust, and those are the pages to rewrite first.

#Step 1: Rewrite your best-ranking pages as answers

Put the direct answer at the top, name the people behind the work and link to the evidence. Keep the foundations that serve both systems: fast pages, clean structure, a sitemap and no crawler blocks.

#Step 2: Answer the follow-up questions

List the questions a buyer asks after the first one, such as cost, timescale, alternatives and risks, and make sure a page on your site answers each plainly. That is how you get cited through query fan-out.

#Step 3: Earn coverage instead of buying links

Stop paying for link volume and put the money into coverage that says something specific about you, in titles your buyers read.

#Step 4: Measure both lines

Keep rankings and visits on one line and mentions on another. Google's Search Console added generative AI performance reports for AI Overviews and AI Mode in June 2026, available to all sites by 31 August 2026. For the other assistants, run a fixed list of 20 to 50 buyer questions several times a month in a clean session, and count a win only when it repeats. ID-Pal's measurement works the same way: "The same buyer questions run every month in ChatGPT, Gemini and Claude, in a fresh chat. A win only counts when it repeats, and losses are reported too."

#What to do this month

1. Verify your site in Bing Webmaster Tools, submit your sitemap and check that robots.txt and your CDN allow Bingbot and OAI-SearchBot.
2. Open your monthly search report and look for the symptoms: stable positions, falling visits, lower click-through.
3. Rewrite the first hundred words of your three highest-ranking service pages so they state plainly what you do, for whom, where and at what size.
4. Build the table above for twenty buyer questions, and note who is named where you are not.
5. List the people in the firm whose names should be attached to its expertise, and check what the web currently says about each.
6. Move any spend that exists purely to buy links towards one piece of coverage in a title your buyers read.

#Frequently asked questions

#Should we keep paying for SEO if ChatGPT ignores rankings?

Yes, because ChatGPT does not ignore rankings: ChatGPT search, Copilot and Google's AI features all start from a search index, so being found there is the first step to being named. What to stop is chasing positions for their own sake with keyword-built pages and bought links. Our guide to whether you still need SEO now buyers ask ChatGPT sets out which parts to keep.

#Which pages on my website does ChatGPT actually read?

OAI-SearchBot crawls pages for ChatGPT search, and ChatGPT-User fetches a page when a user's request needs it. Your server logs show which pages each crawler requests. The main AI crawlers do not run JavaScript, so text that only appears after scripts run may never be read. For clicks, look for visits referred from chatgpt.com in your analytics, and note which pages they land on.

#Can one page be written for both Google and the assistants?

Mostly, yes. A page that answers the buyer's question directly, names the people involved and links to evidence tends to do well in both. Drop the habits of ranking-first writing, such as keyword repetition, long introductions and heavy sales language, and the same page serves both.

#We are a small firm. Does this help us or hurt us?

It can help. Being named rewards clarity and specific expertise, which a small firm can have in abundance. A four-partner practice that says plainly what it does and for whom can be named ahead of a national firm whose site says everything and therefore nothing.

#How long does it take to start being named?

Assistants that search the web live, such as ChatGPT search, Perplexity, Copilot and Google's AI features, can reflect a new page or article within weeks; answers drawn from a model's own memory change only when it is retrained, which takes months. ID-Pal's case study records it: "Claude names it for UK fund administrators, on a page rebuilt in August, and links NorthRow to ID-Pal in the same answer." Measure monthly rather than weekly.

A first-place ranking was hard won and it is still worth something: it gets you read. It no longer guarantees you are the firm the assistant names. To see the gap for yourself, book a free Growth Review: 30 minutes with a founder, with your site, two competitors and what Google and ChatGPT say about all three checked beforehand, and three fixes in writing afterwards.

#Sources