"Best X" and "X vs Y": How to Get Into AI-Generated Comparisons

The comparison is no longer a page, it is an answer

For twenty years, the comparison was a page: a "top 10" written by an editor, frozen until its next update, the same for every reader. That format has just changed in nature. When a buyer asks ChatGPT "best [product] for [use case]" or "[brand A] vs [brand B]", the engine does not display an existing comparison: it composes one, on the spot, for that precise question, with the question's own criteria. Two to five names, arguments, a verdict. And this comparison, recomposed every time, is becoming the first one your buyers read.

For a brand, the question is therefore no longer just "which top lists should we be in?", but "what does the AI need in order to put me in its answer?". That answer changed recently, and the numbers show it.

What changed: the end of the listicle shortcut

Until this year, the recipe seemed simple: engines leaned heavily on third-party rankings, so you "just" had to be in them. That route is still useful, but it is narrowing. According to Peec AI data published in late August 2026, the latest version of ChatGPT cites half as many listicles as before (they now account for just 7.8% of citations) and nearly a third fewer comparison pages, because the engine runs fewer secondary searches on "best", "top" and "versus" (Peec AI analysis). What keeps its citations, per the same analysis: content that answers precisely and directly, anticipating the follow-up questions.

In other words, the engine depends less and less on other people's comparisons: it builds its own from the raw material it finds, product pages, documentation, reviews, discussions. It is what we observe on our audits too: two engines asked the same question on the same day compose different shortlists, from different sources, and the brand that wins is rarely the one dominating the historical "top" lists. It is the one whose facts are easiest to compare.

How the AI composes a comparison

Three ingredients go into the answer, and each one can be worked on.

Comparable facts. To decide between options, the engine needs alignable characteristics: price, capabilities, limits, "who it is for". It looks for them first where they are explicit and in plain text. A brand whose facts are buried in visuals or scattered is hard to place in a table; the engine leaves it out or, worse, fills the gaps with outdated information from its training memory.

Corroborated judgments. The verdict ("better for beginners", "best value for money") comes from the sources that judge: customer reviews, practitioner discussions, press tests, rated directories. Across 680 million citations analyzed by Profound, Wikipedia, Reddit, YouTube and G2 dominate the engines' sources (Profound study): places where people compare and rate, not brochures.

The framing of the question. "Best CRM" and "best CRM for a 20-person SMB" do not produce the same shortlist. The more qualifiers the question carries, the more the engine looks for sources able to answer it precisely, and the better the chances of a brand with an explicit positioning ("built for X, not for Y") against a generic leader. It is the big brands' blind spot and the challengers' window, the same mechanism as on considered purchases.

The action plan in 4 levers

1. Publish your own comparison pages, honest ones

"[You] vs [competitor]", "alternatives to [leader]": if you have no page that answers, the engines will compose with other people's sources. The rule that makes the difference: a factual comparison that acknowledges everyone's strengths gets picked up; a one-sided pitch gets ignored, because the engine cross-checks with other sources and detects the bias. Spell out the use cases where the competitor is the better choice: it is counter-intuitive, and it is precisely what makes the page citable.

2. Make your facts alignable

Turn your product pages into comparison raw material: characteristics in plain text, explicit prices and limits, a "who it is for, what it is for" section that uses your buyers' real qualifiers (size, budget, industry, level). The simple test: if an intern cannot fill in a comparison table in 10 minutes using your site alone, neither can an AI. The detailed moves are in the checklist.

3. Exist where the judging happens

Opinion sources remain the second ingredient of the verdict: customer reviews (volume, rating, recency), your category's directories (G2 and equivalents in SaaS, review platforms in B2C), practitioner discussions, press tests. Third-party listicles have not disappeared, their weight has dropped: treat them as one lever among several, no longer the lever. And everywhere, the same requirement: accurate facts, because every error an engine repeats costs you entire comparisons.

4. Measure your comparisons like a ranking

Your strategic "best X for Y" and "you vs [rivals]" questions, tested across several engines, repeatedly: who gets cited, in what order, with which arguments, from which sources. It is a competitive ranking in its own right, measured with the GEO KPIs (citation rate, share of voice), and it moves with every model version: the shift documented by Peec AI this summer reshuffled entire shortlists from one month to the next. Only repeated measurement counts.

B2C, B2B: the same rules, different judges

For a consumer brand, the judges are reviews, press tests and buying guides; the questions revolve around use case and budget. For a SaaS or B2B service, the judges are software directories, "alternatives to" pages and practitioner feedback; the questions revolve around team size, industry and integration, and they are asked by the person writing the shortlist. In both cases, being absent from the generated comparison never shows in your analytics: the buyer clicks nowhere, they simply remember two or three names, and yours is not among them. Strong-image brands share the same blind spot.

Frequently asked questions

Do AIs really make product comparisons? Yes. On "best [product] for [use case]" or "[A] vs [B]" questions, ChatGPT, Gemini, Perplexity and Google AI Overviews compose a comparative answer with two to five named brands, arguments and often a verdict, recomposed for every question.

How does the AI pick the products it compares? By combining its training memory, comparable facts found online (characteristics, prices, positioning in plain text) and corroborated judgments (reviews, directories, discussions, tests). The composition varies with the engine, the phrasing and the moment.

Is being in third-party "top 10" lists enough? Less and less. Peec AI's August 2026 data shows ChatGPT cites half as many listicles as before. Third-party tops remain useful, but the decisive raw material is becoming your own comparable facts and the review sources.

Should we publish a "we vs competitor" page? Yes, factual and honest, acknowledging the cases where the competitor is the better choice. That is what makes it credible and citable; a one-sided pitch gets discarded by corroboration.

How do I know which generated comparisons my brand appears in? By testing your strategic comparison questions across several engines, repeatedly to smooth out variability, and recording who gets cited and from which sources. An automated audit takes that snapshot in minutes across 8 engines.

Can a challenger beat the leader in these answers? Yes, on qualified questions. The more precise the question (use case, size, budget), the more the engine favors the brand whose positioning explicitly answers that case, even against a bigger but generic player.

Conclusion

The comparison that decides purchases is no longer written by an editor once a year: it is generated on demand, question by question, and it recomposes with every engine update. You no longer get in just by appearing in other people's top lists; you get in by becoming easy to compare: alignable facts, an explicit positioning, third-party judges that corroborate, and continuous measurement to know where you stand. This topic moves with every model version, and we will keep this article updated. First step: know which generated comparisons you appear in today, and which ones are being composed without you. The free Sapian Metrics audit shows you in minutes, across 8 engines, against your real competitors.