Why Do AI Tools Recommend Competitors, and How Do I Get Listed?
Recommendation Answers Are Decided Off Your Site, and the Data Says Where
Short answer
AI tools recommend competitors because recommendation answers are assembled from third-party sources rather than vendor sites. In professional services, third-party listicles accounted for 80.9% of citations against 19.1% for self-published lists, so being in somebody else's roundup beats publishing your own.
1. Why do AI tools recommend competitors instead of my product?
Because a recommendation question fans out to independent sources, and your own site is the weakest possible evidence for the claim that you are the best option. The competitors named are the ones present on those sources.
Two mechanisms combine. Google documents query fan-out for AI Overviews and AI Mode as issuing multiple related searches across subtopics and data sources, so "best tool for X" becomes a set of narrower retrievals. And for the comparative ones, the pages that answer well are lists, reviews and community threads rather than product pages.
The second mechanism is evidential rather than technical. A vendor page asserting that it is the leading option in its category carries almost no weight, because every vendor page in the category makes the same assertion. A model assembling a recommendation reaches for a source with no stake in the answer, which is exactly what your site is not.
That reframes the problem usefully. Being described accurately once you are named is an on-page job. Being named at all is an off-site presence job, and the two are usually owned by different people. Teams that try to solve the second with more content are working on the wrong half.
The diagnostic comes first either way. Ask the recommendation questions your buyers ask, repeatedly, and write down every brand and domain that gets named. Oppira records which competitors and sources appear across the public places those answers draw on, so the list arrives as dated evidence rather than as an impression from one afternoon of prompting.
2. Which third-party sources do AI answers cite for recommendation questions?
Independent listicles, review platforms, Reddit and YouTube carry most of them. Third-party listicles took 80.9% of professional services citations, and review sites rose from 1.51% of citations at awareness to 24.27% at the decision stage.
| Source type | What the citation data shows | What it implies |
|---|---|---|
| Third-party "best tools" listicles | 80.9% of professional services citations went to third-party lists against 19.1% for self-published ones (Wix AI Search Lab, 75,000 answers and over 1 million citations, March 2026) | Pitch the people who write roundups. This is PR and partnerships work, not content work |
| Review and comparison platforms | Trust and review sites were 1.51% of citations at the awareness stage and 24.27% at the intent stage (Seer Interactive, 804,491 responses, 1,926 brands) | Claim G2, Capterra and Trustpilot, then get the first dozen reviews |
| The most-cited domain across ChatGPT, AI Mode, Gemini, Perplexity and AI Overviews (Peec AI, 30 million sources) | Participate genuinely where your buyers already are. Faking it is against Reddit rules and specifically called out by Google | |
| YouTube | The strongest single correlate of AI visibility in the Ahrefs 75,000-brand study, at roughly 0.74 | Short screen recordings of real product output do double duty as sales collateral |
| Wikipedia | Roughly 47.9% of ChatGPT's top-10 source share (Profound, reached only through secondary reporting) | Ignore it. A small company will not pass notability, and pushing an article is a reputational risk |
| Your own site | 12% of AI-cited URLs rank in Google top 10 for the original prompt across 15,000 long-tail queries (Ahrefs) | Own the how-to, definition and limitation questions, where a vendor genuinely is the best source |
The pattern across all six rows is consistency of interest. Sources with nothing to gain from the recommendation get cited for the recommendation, and sources with something to gain get cited for the mechanics. Write for the second and earn your way into the first.
3. Do review site profiles change how often a brand gets recommended?
Substantially, and the jump happens early. Seer Interactive found brands with no Trustpilot profile had a median 1% citation rate, while brands with as few as 1 to 13 reviews reached 53.5%.
That is the largest step change in any dataset in this guide, and it is also the cheapest to act on. Going from no profile to a profile with a dozen genuine reviews is a week of work for most companies, and the sample behind it is 804,491 AI responses across 1,926 brands.
Perplexity is the engine where this shows up most directly. Peec AI reported that Perplexity leans on Reddit, LinkedIn and G2 specifically for B2B queries, which fits Perplexity being the engine with the closest behaviour to classic search: 28.6% of its cited URLs rank in Google top 10 for the original prompt, the highest overlap measured.
Practically, that means completeness matters as much as existence. A claimed profile with a description, categories, screenshots and a dozen reviews can rank for "best X" queries. An empty shell created to tick a box cannot, and will not be cited.
4. Do backlinks or unlinked brand mentions matter more?
Unlinked brand mentions, by roughly three times. Ahrefs, across 75,000 brands, found branded web mentions correlated 0.66 to 0.71 with AI visibility against 0.22 for backlinks.
| Factor | Spearman correlation | What it implies |
|---|---|---|
| YouTube mentions | Roughly 0.74 | Strongest factor tested, with an obvious confound: brands large enough to have YouTube presence are large enough to be cited anyway |
| Branded web mentions | 0.66 to 0.71 | Being talked about, with or without a link, is the closest thing to a lever in this table |
| Branded anchors | 0.53, rising to 0.63 for AI Mode | Links whose anchor text is your brand name, rather than links in general |
| Brand search volume | 0.39 | People searching your name directly, which is mostly downstream of everything above it |
| Domain Rating | 0.27 to 0.33 | Classic authority still counts, at roughly half the strength of brand mentions |
| Backlinks | 0.22 | The weakest link factor. A link-building programme is not an AI visibility programme |
| Number of pages on the site | 0.17 to 0.19 | Almost no relationship. Publishing more pages was the weakest factor Ahrefs tested |
The last row is the one worth sitting with, because it contradicts how most content teams are measured. If page count correlates at 0.17 and branded mentions at 0.66 to 0.71, then thirty more articles is a weaker move than getting quoted in three newsletters, and the second is usually cheaper.
5. How do I get listed in third-party "best tools" articles?
Find the roundups that already get cited for your category, work out what each one requires to include a product, then give the author something checkable rather than a pitch.
Five steps. Step one is the part teams skip, and it is what turns this from cold outreach into a shortlist.
- Find which lists are actually being cited. Ask the recommendation questions in each assistant and record every URL cited, not just the brands named. The result is a list of ten to thirty pages that currently decide the answer for your category, and most of them will not be the ones you expected.
- Work out what each list needs to include a product. Some roundups require a free trial, some a minimum review count on a specific platform, some a filled-in vendor profile, and some just an email. The requirement is usually visible in what the existing entries have in common.
- Fix the prerequisites before pitching. Claim and complete your G2, Capterra and Trustpilot profiles, gather the first genuine reviews, and publish a pricing page with real numbers. Reviewers exclude products they cannot evaluate, and missing pricing is the most common reason.Seer found brands with 1 to 13 reviews reached a 53.5% citation rate against 1% for brands with no profile, so this step pays for itself before any outreach happens.
- Give the author something checkable. Send a short note with a working trial link, a screenshot of real output, one measured number with its method, and the honest limits of the product. Reviewers publish specifics, and a claim they can verify is easier to include than a claim they cannot.
- Track inclusion as a metric, not as a favour. Keep a dated list of the roundups you appear in and the ones you do not, and recheck quarterly. Lists are edited, products get dropped, and the citation follows the current version of the page rather than the one that included you last year.
Two adjacent activities belong to the same job and are usually easier than roundup outreach: being quoted in other people's articles, and appearing on podcasts and in newsletters. Both produce branded mentions, which is the factor that correlated at 0.66 to 0.71, and neither requires anyone to add you to a comparison table.
6. Are my own comparison and alternatives pages worth writing?
Yes, but not for recommendation questions. Self-published lists took 19.1% of professional services citations against 80.9% for third-party lists, so your own pages win narrow branded comparisons rather than the category question.
The useful split is by who is asking. Somebody typing "best competitive intelligence tools" is early and will be served third-party lists. Somebody typing "product A versus product B" has already narrowed, and a comparison page that treats the other product fairly can be the best available source for that specific pair.
So write the pages that answer questions you are genuinely the best source for: how your product handles a specific job, what it does not do, what data it can and cannot access, how it differs from one named alternative. Those are answerable honestly by a vendor, and they are the questions where a vendor page is not automatically discounted.
What does not work is a self-published "top 10 tools" article that ranks your own product first. It competes directly with independent lists that hold 80.9% of the citations for that intent, on the one dimension where you have no credibility. The effort is better spent getting into three of those lists.
7. Which off-page tactics do not work?
Buying mentions, astroturfing community threads, chasing a Wikipedia article, and publishing more pages. Google explicitly names inauthentic mentions as ineffective, and page count was the weakest factor in the Ahrefs study.
Four more things worth deliberately not doing:
- Chasing a Wikipedia article. Wikipedia is roughly 47.9% of ChatGPT's top-10 source share, and a small company will not pass notability. A rejected or deleted article is a public record of the attempt.
- Treating a Wikidata item as an AI visibility tactic. It is worth a few hours for entity disambiguation and sameAs wiring, and no published study isolates any effect on citation. Do it, expect nothing measurable.
- Publishing more guides as the plan. Number of site pages correlated 0.17 to 0.19 with AI visibility, the weakest factor Ahrefs tested across 75,000 brands.
- Creating a separate page for every query variation. Google calls that both a spam policy concern and an ineffective long-term strategy in its own AI optimisation guidance.
The honest summary is that the effective off-page work is slow and unglamorous: real reviews, real participation, real mentions in things other people publish. It cannot be automated, which is precisely why it still differentiates.
Key Takeaways
Third-party lists carry recommendation answers
In professional services, third-party listicles took 80.9% of citations against 19.1% for self-published lists across over 1 million citations (Wix, March 2026).
A review profile with a dozen reviews is a step change
Seer found a median 1% citation rate for brands with no Trustpilot profile and 53.5% for brands with as few as 1 to 13 reviews.
Review sites matter at the decision stage, not at awareness
Trust and review sites were 1.51% of citations at the awareness stage and 24.27% at the intent and decision stage.
Unlinked mentions beat backlinks by roughly three times
Branded web mentions correlated 0.66 to 0.71 with AI visibility against 0.22 for backlinks, across 75,000 brands.
Publishing more pages is the weakest lever measured
Number of site pages correlated 0.17 to 0.19 with AI visibility, the lowest of the factors Ahrefs tested.
The mechanism is organic ranking, not AI preference
Seer found 99.5% of Trustpilot citations arrived through organic rankings, so an incomplete profile that cannot rank will not be cited.
Frequently Asked Questions
Sources
- AI citations favor listicles, articles, product pages, reporting the Wix Studio AI Search Lab study Search Engine Land / Wix, March 24, 2026.75,000 AI answers and over 1 million citations across three engines. Vendor study reported by trade press, with the full methodology not published.
- Study of 800K AI Responses: How Review Profiles Shape Brand Presence Seer Interactive, January 2026.804,491 responses across 1,926 brands. Agency research with disclosed method. Also the source of the finding that 99.5% of those citations came through organic rankings.
- Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands) Ahrefs, January 2026.Spearman correlations with a Domain Rating above 40 filter. The authors explicitly warn that correlation is not causation and that all correlations are moderate to weak.
- AI search engines cite Reddit, YouTube and LinkedIn most, reporting Peec AI Search Engine Land / Peec AI, March 31, 2026.30 million sources. Trade press summary; exact percentages were not disclosed, which is why the Reddit finding is stated as an ordering rather than a figure.
- Only 12% of AI Cited URLs Rank in Google's Top 10 Ahrefs, January 2026.15,000 long-tail queries across ChatGPT, Gemini, Copilot and Perplexity. Source of the per-engine overlap figures including Perplexity at 28.6%.
- Optimizing your website for generative AI features on Google Search Google, July 10, 2026.Primary. Names buying inauthentic mentions among the things that do not work, and calls separate content for every query variation an ineffective long-term strategy.
- Spam policies for Google web search Google, July 2026.Primary. The definition of scaled content abuse behind the guidance against creating a page for every query variation.
- AI features and your website Google, July 2026.Primary. Documents query fan-out as issuing multiple related searches across subtopics and data sources.
- What Drives Brand Mentions in AI Answers? Seer Interactive, January 2025.300,000 keywords in finance and SaaS. The authors call it phase one and flag that correlation is not causation.
Explore More
Related analyses, benchmarks, and industry insights
Related Guides
Glossary Terms
Turn reading into a reaction
Oppira watches your competitors, keeps your playbook current, and drafts the response. Start free and see your market clearly by tomorrow.