Why does AI recommend my competitors and not me?

Pick the prompt that hurts most, usually the category question where the rival is named first. Ask it in a temporary chat with search on, in each assistant your buyers use, and open every source the answer attaches. For each page, note whether it names the rival, whether it names you, and how much of the page each of you gets. The pattern is usually visible in one sitting: the rival is on four of the five cited pages and you are on one, or on none.

This is not a quirk of one engine. Independent analyses of AI citations put the weight on third-party pages: around 84 per cent of AI citations come from earned media (Muck Rack, from an analysis of more than 25 million AI-cited links, 2026), and around 63 per cent of B2B 'best tools' citations point to listicles and roundups (Evertune, across 25,000 URLs). Both figures, with their sources, are on our GEO statistics page. Your own site is one voice in that set, and the one with the most obvious motive.

Depth first. On the cited pages, a winning rival tends to get a paragraph: what it does, who it suits, a price, a switching story from a named use case. The losing brand gets a clause in a list, or a row in a table with no description at all. Copy the rival's description from each cited page into a document next to yours, and the difference is usually stark.

Then consistency. Assistants favour brands that many sources agree on, and contradiction costs you the slot even when you are present. If one page calls you a 'revenue platform', another calls you 'invoicing software' and your own site says 'financial operations', the assistant has three entities to reconcile and a rival with one. Check the category label and the core claim on each cited page against your homepage; every mismatch is a fix you control.

The rival is often filed under the phrase the buyer typed and you are not. 'Best CRM for a small accountancy firm' finds the brand that review sites and listicles describe with those words. If third parties describe you with the words on your homepage rather than the words your buyers use, the assistant has to bridge the gap on its own, and it usually does not. Read the rival's category wording on the cited pages, then read yours; where they differ, the rival's wording is the one to adopt on your own site and in the listings you can edit.

Assistants lift written comparison from review pages, and crawlers see what the page shows first, which is the newest reviews. Open the rival's profile and yours on the review sites your category uses (G2 and Capterra for software, Trustpilot for consumer services). Note the date of the newest review, the count in the last six months, and how many reviews name a use case and a product they switched from. A rival with twelve reviews this quarter that read like buying advice will be quoted ahead of a brand with a higher star average and nothing newer than 2024. Do AI assistants read your reviews? covers what gets quoted and why.

Engines expand a prompt into sub-questions before they answer, so a rival can win the recommendation on questions you never asked. 'Best expense software for agencies' becomes questions about pricing, integrations, alternatives and reviews, and the brand present on the pages that answer those sub-questions is the one that reaches the shortlist. The visible traces are Google's People Also Ask box for the query and, in ChatGPT with search on, the searches it ran where the surface shows them. List the sub-questions, then check which brand each cited page answers for. Discoverable's Query fanout maps the sub-questions behind a prompt and scores your presence across them.

Turn the comparison into three numbers per prompt, per engine: whether you are named, your position among the brands named (first, in the list, or an afterthought), and share of voice, meaning how many of the answers for that prompt name you against how many name the rival. Named in 2 of 5 runs against the rival's 5 of 5 is the plain statement of the problem, and it tells you which rival, which prompt and which engine to work on first.

Discoverable produces the same numbers on a schedule. The competitor leaderboard ranks every brand named across your prompts by share of voice per engine, the citation gap counts the sources cited for rivals that never cite you, and the free AI visibility check shows the competitors ahead of you and a ranked citation gap with no login. The free check runs on Claude with live web search; paid plans also cover ChatGPT, Perplexity, Gemini, Copilot, Google AI Overviews and Google AI Mode. For ChatGPT in particular, the ChatGPT visibility tracker records the public answer, the position of every brand it names and the pages it cites.

Three neighbouring guides cover the cases this one does not. If AI never names you at all, the causes are different and ranked in why doesn't ChatGPT mention my brand?. If it used to name you and stopped, start with why has ChatGPT stopped recommending my brand?. If what you want is the full list of pages behind the rival's recommendations, how to find the sources AI uses to recommend your competitors is the procedure. The mechanism underneath all of it, consensus across independent sources, is explained in how AI assistants choose which brands to recommend.