Why ChatGPT Recommends Your Competitors and Not You

You asked ChatGPT for the best vendors in your category. It named your competitors, and not you. The question that follows is a fair one: what did they do that you did not?

Nobody outside the engine can answer that for certain, and anyone who says they can is selling something. What you can do is look at what the answers actually cite, check whether the engines can reach your pages, and test one change at a time. This page shows what we saw when we checked 43 answers to software-buyer questions, what OpenAI, Perplexity and Google say about how sites get included, and five checks to run before you spend money.

It is written for marketing leads and founders at B2B software and developer-tool companies. It describes one day of our own data and a small number of published sources. If the term GEO is new to you, start with What Is Generative Engine Optimisation. It does not promise a placement. Our methodology page explains how we run the panel.

Why It Feels Unfair, and Why the Answer Is Not Fixed

When a buyer asks an assistant for "the best tool for X", the assistant does not return a ranked list from a single index. With web search on, it runs one or more searches, reads pages and writes an answer. Which companies appear depends on what it found and how it weighed those pages.

That process is not stable. A study by SparkToro and Gumshoe, reported by Search Engine Journal on 30 January 2026, ran 2,961 prompts across ChatGPT, Claude and Google's AI features. It found that repeated runs of the same prompt almost never produced the same brand list. The same coverage notes that the volunteers used their normal settings and the report is not peer-reviewed. The original study is worth reading in full.

So one answer that omits you is a single draw. It tells you something, but less than it feels like. We explain how to turn many answers into a range in Why an AI Visibility Score Needs a Confidence Interval.

What Software Buyers Say They Do

The reason this matters is buyer behaviour. In March 2026, G2 surveyed 1,076 B2B software buyers and decision-makers. In its published research, 51% said they start research with an AI chatbot more often than with Google, up from 29% in G2's 2025 report. G2 also reports that 69% of buyers chose a different vendor than they had planned because the vendor was in a chatbot's recommendation.

Read those figures with care. G2 sells review and marketing products to software vendors, so it has an interest in the story. The survey covers software buyers only, so it says nothing about other industries. And the 69% describes what buyers reported, not what we or anyone measured about the chatbot's effect. Even so, if your buyers behave like that sample, an answer that omits you is a missed shortlist.

What the Answers Actually Cite

We wanted to see what sits behind the omissions. On 18 October 2026 we asked nine buyer-style questions for B2B software companies on five AI assistants, each question once: ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, each with web search. Examples include "Why does ChatGPT recommend my competitors' software but not ours and how do we fix it?" and "Best way for a B2B SaaS company to appear in best-software-for-X answers from AI assistants". That is 45 requests, and 43 returned an answer.

None of the 43 answers named ThriveFinity. With 43 answers, that is consistent with a true rate of up to 8.2%, so we read it as not seen yet, not never.

More useful is what the answers cited. Thirty-seven of the 43 included at least one source link. Counting the answers that cited each site, the most common sources were:

Two things stand out, and we hold both loosely. First, the cited pages were mostly third-party or neutral: a social network, help pages from the engines themselves, a video site, and blogs from agencies that sell this work. Second, no single vendor site dominated. The most-cited single site appeared in 8 of 43 answers, which is under one in five.

What this does not show is why those pages were chosen, or that appearing on them would change an answer. It shows where the engines looked on one day. If your competitors are named in answers that cite pages you are absent from, that is an association you can test. It is not a cause you can assume.

Not Every Assistant Shows Its Sources the Same Way

The split by assistant matters for how you check. In our run, every ChatGPT, Claude and Perplexity answer to these nine prompts included at least one source link (9 of 9 each), and so did all 7 answers that Google AI Overviews returned. Gemini included a link in 3 of its 9 answers.

We would not read that as "Gemini uses fewer sources". Link capture depends on how each product returns its citations to our tooling, so a missing link may mean the data came back without one, not that the answer used no sources. It does mean that an audit built only on citations will see some assistants more clearly than others. For the ones that show less, read the answer text as well, and note which companies and which claims it repeats.

Two of the 45 requests, both to Google AI Overviews, failed with a data-provider error. We report them and leave them out of the 43 answers, which is the same rule we use in every run.

The practical point for a software company is to keep the check simple and repeatable. Use the same questions each week, record the same fields, and treat a difference between assistants as a lead until it shows up more than once.

Can the Engines Reach You at All?

The first check is the dull one, and it is often missed. Each product has its own crawler and its own rules.

OpenAI's ChatGPT search help page says a website can be included in ChatGPT search if it allows OAI-SearchBot to crawl it and its host or CDN permits traffic from OpenAI's published searchbot IP addresses. It adds that inclusion does not guarantee a particular ranking. A site that blocks that crawler in robots.txt, or sits behind a firewall rule that drops the traffic, has given itself no chance.

Perplexity's crawler documentation says PerplexityBot exists to help sites appear and be linked in Perplexity's results, and that it is not used to gather data for building foundation models. It recommends allowing the bot in robots.txt.

Google's documentation on AI features says a page qualifies as a supporting link in AI Overviews and AI Mode if it is indexed and eligible to appear in Search with a snippet, with no additional technical requirements. It also says that being eligible does not guarantee a page will be crawled, indexed or shown.

None of this is exotic. It is the same crawlability discipline you already apply for search, extended to a few more user agents. We cover the practical side in our guide to running an AI visibility check on your company.

Five Checks Before You Change Anything

1. Check access. Confirm your robots.txt and CDN allow OAI-SearchBot, PerplexityBot and Googlebot, and that nothing returns an error to them. Fix this first because it costs little.
2. Check your own answers. Run your real buyer questions on at least three assistants with web search on. Record who is named and which pages are cited. A dozen questions is a start, but it is a small sample, so repeat the run before you draw a conclusion.
3. Read the cited pages. For each answer that names a competitor, open the pages it cites. Ask what those pages say that yours does not: a comparison, a price, a use case, a named customer type, a date.
4. Look at where you are absent. If answers cite third-party guides, review sites, video explainers or community threads, list which of them mention you accurately. Our check suggests answers draw on pages outside a company's own site.
5. Change one thing and re-run. Publish or fix one page, wait, and run the same locked prompts again. Judge the result with ranges, as set out in our share-of-voice guide, not by a single before and after answer.

If you sell software, a page that states plainly who the product is for, what it costs, what it does not do and how it compares on named criteria tends to be easier for anyone to cite, human or machine. That is a hypothesis from the pages we saw cited, not a proven rule, and we will report on it as our weekly runs build up.

What Nobody Can Promise

Treat any promise of a guaranteed placement in an AI answer as a warning sign. The engines change their models, sources and ranking logic without notice, and OpenAI states in its own documentation that inclusion does not guarantee a ranking. What a good agency or tool can do is measure honestly, find the gaps and fix what is within your control. Our buyer's guide to choosing a GEO agency lists the questions to ask and the red flags to watch for.

If you want a repeated measurement with ranges, written up by a named person, that is what the AI Visibility service provides. For software companies specifically, see our page for B2B SaaS teams.

Limits of This Page

• The cited-domain counts come from one day, one panel, nine prompts and one pass per prompt and surface.
• Citations show where an engine looked. They do not show why it chose a page, or that a page changed the answer.
• The G2 survey is a vendor survey of software buyers in one month. It does not cover other industries.
• OpenAI, Perplexity and Google can change their documentation. The statements above were checked on 19 October 2026.
• We name no vendor scores. Any error we find is logged on the errata page.

Spotted something out of date or wrong on this page? Email om@thrivefinity.uk and we respond within 5 working days.

Pranav Unni

Founder · ThriveFinity Connect on LinkedIn →

Pranav Unni is the founder and lead verifier of ThriveFinity. He reads and signs every paid deliverable personally, and writes about go-to-market decisions for established B2B companies.