# How do we get ChatGPT to recommend our firm when buyers ask?

Canonical: https://fortitudemedia.ai/insights/what-is-llm-optimisation
Author: David Adams
Published: 2026-04-22
Last reviewed: 2026-09-24
Reviewed by: David Adams
Pillar: Getting named by AI assistants
Summary: How ChatGPT, Claude, Gemini and Perplexity decide which firms to recommend, what a B2B firm can change, and how to check whether you are being named.
Keywords: how to get recommended by chatgpt, get my business on chatgpt, why does chatgpt recommend my competitor, can you pay to be recommended by chatgpt, ai search b2b
Short answer: Give the assistants three things: credible mentions of your firm on other sites, a website that says plainly what you do and for whom, and visible evidence of expertise. Assistants find firms from training data, which changes over months, and live search, which can change within weeks: ChatGPT search and Copilot draw on Bing, Google's AI features on Google. Ranking helps but does not guarantee a mention. Track 20 to 50 buyer questions monthly and count a win only when it repeats.
What changed: Reworked around how assistants choose firms: training data versus live search, why competitors get named, paid ads in ChatGPT, reach figures, and the ID-Pal and Helm Club results.
Publisher: Fortitude Media Limited

When quoting, attribute to Fortitude Media and link to the canonical URL above.

---
For twenty years the question was "where do we rank on Google?" You paid an agency, they reported positions, and if you were on page one for the right terms the enquiries came. Then a prospect mentioned, almost in passing, that they had asked ChatGPT for a shortlist before they rang anyone, and your firm was not on it. You checked. They were right.

That is the change in one sentence. Buyers ask ChatGPT, Gemini, Claude, Copilot and Perplexity who to hire, and the assistants answer with names. The audience is large: OpenAI said on [31 August 2026](https://openai.com/index/expanding-access-to-ai-with-chatgpt-ads/) that ChatGPT has more than a billion weekly users, and [Forrester found in January 2026](https://www.forrester.com/blogs/b2b_buyers_make_zero_click_buying_number_one/) that 94% of business buyers use AI somewhere in their buying process.

Being one of the names builds on search but is not the same job. This guide explains what an assistant is doing when it names a firm, the two routes by which your firm gets into an answer, what the assistants look for, why a competitor gets named instead of you, and how to tell whether your work is paying off.

## Key takeaways

- An AI assistant answers a buyer with a handful of names, so a firm that is left out is invisible for that question.
- Firms get into answers by two routes: what the model learned in training, which changes over months, and live web search, which can change within weeks.
- ChatGPT search and Copilot draw on Bing's index, while Gemini, AI Overviews and AI Mode draw on Google's, so ranking in both helps but does not guarantee a mention.
- Assistants name the firms they are most confident about: firms mentioned credibly on other sites, with a website that says plainly what they do and visible evidence of expertise.
- Nobody can buy a place inside the answer; ChatGPT has shown labelled ads to UK users on its Free and Go plans since 11 August 2026, and OpenAI says the ads do not change the answer.
- Measure progress with a fixed list of 20 to 50 buyer questions, asked several times a month, counting a win only when it repeats.

## What is an AI assistant doing when it names a firm?

It reads what it already knows and, increasingly, what it can find by searching, then writes an answer naming the firms it is most confident about.

ChatGPT, Claude and Gemini are built on large language models: systems trained on vast amounts of text so they can understand a question and answer it in plain language. Perplexity is an answer engine that searches the web for each question and writes the answer with a language model. None of them returns ten links. They read the question, weigh the evidence and write an answer that names specific firms.

Ask "which employment law firms should a manufacturing business in the Midlands talk to" and the answer names a few firms; how many varies by assistant, wording and user. If yours is among them, you have a recommendation from an intermediary the buyer trusts.

## Two routes into an answer: memory and live search

An assistant can know about your firm because it learned about you in training, or because it found you by searching while it wrote the answer. The two routes work on very different timescales.

| Route | How it works | Which assistants | How fast it changes |
|---|---|---|---|
| Memory (training data) | The model learned about firms from text gathered before its training cut-off | All of them, when they answer without searching | Only when the model is retrained, which takes months |
| Live search | The assistant searches an index, reads the top pages and cites them | ChatGPT search (which has drawn on Bing), Copilot (Bing), Gemini, AI Overviews and AI Mode (Google), Perplexity and Claude, which search the web | Within weeks of a new page or article |

For live search, your pages have to be in the right index and open to the right crawler: Bing for ChatGPT search and Copilot, Google for Google's AI features, and OpenAI's OAI-SearchBot, Anthropic's Claude-SearchBot and PerplexityBot allowed in. [How AI crawlers read your website](/insights/ai-crawlers-vs-google-spiders) shows how to check. For memory, what matters is how often and how consistently your firm was described across the web before the last training run, which changes only as others write about you.

When an assistant does search, a clearer page or a new article can show up in its answers within weeks.

## How is being named different from ranking on Google?

Google's classic results make you findable; an AI answer endorses a few firms. Ranking helps, but it does not guarantee a mention.

Google's AI Overviews and AI Mode, ChatGPT search and Copilot all start from a search index, so a page that ranks is more likely to be read and cited. But the link is loose. [Ahrefs found in March 2026](https://ahrefs.com/blog/ai-overview-citations-top-10) that only 38% of the pages cited in AI Overviews ranked in the top ten for the search, down from 76% in July 2025, because Google now runs several related searches for one question (query fan-out). An assistant looking at a buyer's question also reads pages that answer the follow-ups: what it costs, how long it takes, who else does it.

So covering the follow-up questions matters as much as ranking for the main one. [We rank first on Google, so why does ChatGPT not mention us?](/insights/google-rankings-wont-protect-you) covers that gap, and [our comparison of SEO and AI search](/insights/seo-vs-llm-optimisation) sets out which SEO work to keep, stop and add.

## What does an assistant look for before it names you?

Three things: credible mentions of your firm elsewhere, a website that says plainly what you do, and visible evidence of expertise. Assistants are weighing patterns in what they read, and these three patterns decide most of it.

### Credible mentions on other sites

An assistant is far more confident naming a firm that independent sources discuss. Muck Rack's [May 2026 analysis](https://www.globenewswire.com/news-release/2026/05/07/3290268/0/en/generative-pulse-earned-media-consistently-drives-ai-citations-holding-at-84.html) of more than 25 million links cited by ChatGPT, Claude and Gemini found 84% came from earned media, meaning coverage by others, and 0.3% from paid content. Review sites count too: in [G2's April 2026 survey](https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html) of software buyers, 45% said citations of review sites were the most confidence-inspiring signal in an AI answer. Trade press, professional directories, awards, partner pages and client case studies all tell the assistant that others have checked you. Brand recognition built this way helps with assistants as well as with Google. [Which kinds of PR win work for a B2B firm](/insights/pr-for-b2b-what-works) covers how to earn it without a large budget.

### A website that says plainly what you do

If your description could apply to a hundred other firms, your services are three clicks deep and your expertise is implied, the assistant cannot confidently say what you do or who you are for. Say it in the first screen of the homepage and every service page: what you do, for whom, where and why you are different. Structured data, a clear heading hierarchy and a logical site structure make that easier for a machine to read; [whether you need schema markup](/insights/structured-data-schema-markup-ai) explains what it does and does not do.

### Visible evidence of expertise

If your people have published considered work on your subject, answered the hard questions and named themselves as the authors, the assistant can see it. Thin, generic content does the opposite. Real examples, specific figures with sources and plain discussion of trade-offs are what separate your pages from the rest. [What E-E-A-T is and why AI cares](/insights/eeat-why-ai-cares-more) goes into how assistants judge this.

## Why does ChatGPT recommend my competitor and not me?

Usually because the competitor gives the assistant more to work with on the specific question asked. Work through the likely causes in this order, because each makes the next irrelevant until it is fixed.

| Likely cause | How to check | What to do |
|---|---|---|
| The assistant cannot reach or find your pages | Check robots.txt and CDN settings for blocked AI crawlers; search Bing for your key pages | Allow the search crawlers; verify your site in Bing Webmaster Tools |
| Your site does not say plainly what you do | Read your homepage as a stranger for thirty seconds | Rewrite the first screen of the homepage and each service page |
| Nobody else writes about you | List every third-party mention: press, directories, reviews, partners | Earn a few credible mentions in the places your buyers read |
| The competitor answers the follow-up questions and you do not | Compare their guides and service pages with yours on cost, timing and fit | Write the page that answers each follow-up completely |
| Sources disagree about you | Compare your name, services and founding facts across your site, LinkedIn and directories | Make every profile say the same thing |

Old search tactics can also count against you: hundreds of near-identical keyword pages read as a site built to game a system. [Why ChatGPT ignores sites with hundreds of pages](/insights/ai-penalises-thin-content) covers the fix, and [why ChatGPT describes a firm wrongly](/insights/ai-conflicting-information-online) covers inconsistent facts.

## Does ChatGPT give different recommendations to different people?

Yes. Answers vary from one run to the next, with the exact wording of the question, and with what the assistant knows about the person asking, such as their location or, where it is switched on, ChatGPT's memory of earlier chats.

That is why a single test proves little. A founder who asks once and sees their firm named, or not named, has learned very little. What counts is how often you are named across a fixed set of questions asked repeatedly in clean sessions.

## Which matters more for B2B: ChatGPT or Google AI Overviews?

Both, in that order, then Copilot and Gemini, then Perplexity and Claude. The reach figures explain why.

| Assistant | Reach | Why it matters for B2B |
|---|---|---|
| ChatGPT | More than a billion weekly users (OpenAI, August 2026) | The standalone assistant with the largest audience |
| Google AI Overviews and AI Mode | AI Overviews over 2.5 billion monthly users, AI Mode over 1 billion ([Alphabet, June 2026](https://blog.google/alphabet/investor-presentation-june-2026/)) | Shown inside ordinary Google searches, so buyers see it without choosing to |
| Microsoft Copilot | Microsoft 365 Copilot passed 30 million paid business seats ([Microsoft, July 2026](https://www.microsoft.com/en-us/investor/earnings/fy-2026-q4/press-release-webcast)) | Built into the tools many buyers use at work; Forrester found business buyers four times as likely as consumers to use Copilot |
| Gemini | Gemini app over 900 million monthly users (Alphabet, June 2026) | Draws on Google Search, so Google work carries over |
| Perplexity and Claude | Smaller audiences | Both search the web, so the same work covers them |

Because ChatGPT search and Copilot lean on Bing and Google's features lean on Google, one programme covers most of them. [What Google AI Mode is and how it differs from AI Overviews](/insights/end-of-ten-blue-links) covers Google's side, and [which sources ChatGPT cites when a B2B buyer asks who to hire](/insights/ai-optimisation-b2b-vs-b2c) covers the B2B pattern.

## Can we pay to get our firm recommended by ChatGPT?

Not inside the answer. ChatGPT has shown labelled ads to UK users on its Free and Go plans [since 11 August 2026](https://openai.com/index/testing-ads-in-chatgpt/), and OpenAI says the ads do not change the answer itself. Plus, Pro, Business, Enterprise and Education plans show no ads. Being named in the answer is still earned, which is why the three things above matter more than any ad budget.

Be wary of anyone who says they can guarantee a place in ChatGPT's answers. Nobody can guarantee what an AI assistant will say. A provider can guarantee effort, reporting and outcomes it controls, such as pages published or meetings held.

## How do we know whether it is working?

Track a fixed list of 20 to 50 questions your buyers ask, in their words. Put each to ChatGPT, Gemini, Claude, Copilot, Perplexity and Google's AI Mode several times a month in a clean session, because answers vary from run to run, and count a win only when it repeats. Record whether you are named, where in the answer, and how you are described. [How to find out whether ChatGPT recommends your firm](/insights/track-ai-recommending-your-business) shows how to do this by hand, and our [example monthly report](/samples/monthly-report) shows one way to present it.

Two of our clients show what the results look like when the method is kept up.

[ID-Pal](/work/id-pal), an identity verification firm, started from a familiar place: "AI could describe ID-Pal. It never recommended it." Its measurement: "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." The result so far: "UK buyer questions where AI names ID-Pal: none in May, two in August, four in September 2026."

[Helm Club](/work/helm-club) runs chaired peer groups for founders of growing UK businesses. "A channel that did not exist before now brings one in five new members: founders who ask ChatGPT, Gemini or Claude which peer group to join are told about Helm." Its reporting is kept to what the business can check: "Qualified calls a week, reported on one page every Monday and checked against Helm's own member records rather than a platform dashboard."

Alongside the question list, ask every new enquiry how they found you, because many buyers read an answer and then search your name.

## What does an effective programme look like?

Three things working together: a website built to be understood, content that shows real expertise, and a steady programme of earned coverage. Any one alone moves slowly.

The website is the foundation, the content is the proof, and the coverage is the corroboration that lets an assistant check your claims against someone else's account. You can do this yourself with a plan and discipline; for a guide to each part, see [being named by AI assistants](/insights/pillar/ai-llm-optimisation).

## What to do this month

1. Read your homepage as a stranger and check whether the first screen says what you do, for whom and why you are different. Rewrite it if not.
2. Check your robots.txt and CDN settings let OAI-SearchBot, Claude-SearchBot and PerplexityBot in, and verify your site in Bing Webmaster Tools.
3. List every place your firm is mentioned by a third party: press, directories, reviews, client and partner sites. If the list is short, that is your first gap.
4. Find your three most substantive pieces of published content and your thirty thinnest. Plan to strengthen the first and merge or retire the second.
5. Write down 20 to 50 questions your buyers ask and run them through the main assistants to set a baseline.
6. Decide who owns this, because a programme that is nobody's job stops in month two.

## Frequently asked questions

### How much does it cost to get our business recommended by AI?

It depends on how much of the work you already have: a clear website, published expertise and outside coverage. Some firms need only rewritten pages and a few earned mentions; others need a rebuild. [What AI search optimisation costs in the UK](/insights/ai-search-optimisation-cost-uk) sets out the market, and our own prices are on [what it costs](/what-it-costs).

### We are a niche or regional firm. Does this still apply?

Yes, and often more strongly. When someone asks for a specialist in a narrow field or a particular region, the assistant still has to name someone, and the pool is smaller. A firm with clear positioning and a few credible third-party mentions can own a niche question in a way that is impossible for a broad one.

### How do we get third-party mentions without a big PR budget?

Start with what you already have: client case studies with permission, your professional body's directory, sector awards, trade publications that need expert comment, review profiles and partner websites. Five credible mentions in the places your buyers read do more than fifty in the wrong ones.

### How long before we see a change?

Assistants that search the web live (ChatGPT search, Perplexity, Copilot and Google's AI features) can reflect a new page or article within weeks. Answers drawn from a model's memory change only when it is retrained, which takes months. ID-Pal's case study gives a dated example: "UK buyer questions where AI names ID-Pal: none in May, two in August, four in September 2026."

### Do our Google rankings become irrelevant?

No. Rankings feed AI Overviews, AI Mode and, through Bing, ChatGPT search and Copilot, so they still help. They just do not guarantee a mention, and the tactics that once produced them cheaply, such as keyword pages and bought links, do nothing for an assistant.

The firms that get named are rarely the biggest. They are the ones whose expertise is written down, corroborated and easy to read. To see where you stand today, [book a free Growth Review](/growth-review): 30 minutes with David Adams and three fixes in writing.

## Sources

- [OpenAI, A milestone in expanding access to AI (August 2026)](https://openai.com/index/expanding-access-to-ai-with-chatgpt-ads/)
- [OpenAI, Testing ads in ChatGPT (updated August 2026)](https://openai.com/index/testing-ads-in-chatgpt/)
- [Forrester, B2B buyers make zero-click buying number one (January 2026)](https://www.forrester.com/blogs/b2b_buyers_make_zero_click_buying_number_one/)
- [Ahrefs, AI Overview citations from top-ten pages (March 2026)](https://ahrefs.com/blog/ai-overview-citations-top-10)
- [Muck Rack via GlobeNewswire, Earned media consistently drives AI citations (May 2026)](https://www.globenewswire.com/news-release/2026/05/07/3290268/0/en/generative-pulse-earned-media-consistently-drives-ai-citations-holding-at-84.html)
- [G2 via PR Newswire, Half of B2B software buyers now start their research with AI chatbots (April 2026)](https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html)
- [Google, Alphabet investor presentation (June 2026)](https://blog.google/alphabet/investor-presentation-june-2026/)
- [Microsoft, FY26 fourth quarter earnings (July 2026)](https://www.microsoft.com/en-us/investor/earnings/fy-2026-q4/press-release-webcast)