Why AI names your competitors instead of you: a diagnostic checklist

Start by reading the answers

Before diagnosing anything, collect evidence. Run the question several times on each engine and save the full answers and the sources they cite. A single answer is an anecdote; ten runs show the pattern.

Note three things for each run: which brands are named and in what order, what reasons are given for recommending them, and which sources are cited. The reasons and the sources point directly at the cause.

Cause 1: access. Can the engine reach your pages?

The cheapest problem to fix is also the easiest to miss. If your robots.txt blocks the crawlers an engine uses for search, or your pages only render with client-side JavaScript that crawlers do not execute, or your server returns errors to unfamiliar user agents, the engine cannot use you even if you would otherwise qualify.

Check your robots.txt for rules that name AI crawlers, check that key pages return full content to a plain fetch, and review server or CDN logs to see whether AI crawlers are requesting your pages and what response they get. If a crawler never fetched the page, that explains the absence and fixing it is straightforward.

Related reading: How citations shape the generated shortlist and How to write a page AI answers will actually cite.

Cause 2: clarity. Do your pages state the answer?

Engines favour pages that answer a question directly. If the question is 'best invoicing tool for freelancers' and your page talks about 'empowering financial workflows', there is nothing concrete to quote.

Look at the pages the engine cited for this question and compare. They tend to state who the product is for, what it costs, how it compares, and what the limits are, in plain sentences near the top. Write the equivalent: a page that opens with the answer, supports it with specifics and dates, and says honestly where the product is not a fit.

Cause 3: corroboration. Do trusted sources mention you?

For comparison and recommendation questions, engines often lean on third-party sources: review sites, industry publications, forums, directories and listicles. If those sources list your competitors and not you, the answer will too.

List the domains cited across your tracked prompts. Which of them mention you? Which mention competitors but not you? That gap is your outreach and content list: get accurate listings, earn honest reviews, contribute original data that others cite, and correct outdated entries.

Cause 4: association. Does the engine connect you to the category?

Models build a general sense of which brands belong to which categories. If your own pages and the wider web describe you inconsistently, for example as a CRM in one place and a marketing suite in another, the association is weak and you are less likely to be offered for either.

Pick the category language you want to be known for and use it consistently: on your homepage, your product pages, your profiles on directories, your press descriptions and your structured data. Consistency across sources is how a category association forms.

Cause 5: coverage. Are you in the prompts that matter?

Sometimes the problem is narrower than it looks. You may be visible for 'best X for agencies' and invisible for 'best X for ecommerce', because you have no content or evidence for that segment.

Break your prompts into groups by audience, use case and market. Find the groups with zero mentions and check whether you have any page, mention or case study that could support an answer there. Closing a coverage gap often means building a specific page and earning one or two credible third-party references.

Turning the diagnosis into a plan

Rank your findings by cost and by blocking effect. An access problem blocks everything and is usually quick to fix, so it goes first. Clarity fixes are next, since they are fully in your control. Corroboration and association take longer and compound over months.

Then measure again with the same prompts. Keep the prompt set fixed so you can tell whether a change helped, and keep the raw answers so you can read what changed, not only the percentage.

In OClever, the Prompts view shows for each question who was named instead of you, the sources cited, and the sentences used. The Agent Readiness checks cover the access causes, and Content Studio drafts pages aimed at the gaps you find.