For "best X for Y" buyer questions, AI engines lean hardest on one page type: the independent "best of" list. New research puts a number on just how dominant it is, and points to a clear, if uncomfortable, playbook.
How much do best-of lists dominate AI answers?
In a 2026 study, Glen Allsopp analysed around 750 top-of-funnel prompts across ChatGPT (software, products, and agencies) and roughly 26,283 source URLs. The headline finding:
- "Best X" blog lists were 43.8% of all cited page types: the single most-cited format.
- Of 1,100 best-of lists examined, 79.1% had been updated in 2025, and 26% within the prior two months: strong evidence that engines favour recently refreshed comparison content.
- Brands in the top third of a list were more likely to surface in ChatGPT answers than those further down.
- About 35% of cited best lists came from low-authority domains: many built mainly for link-building rather than genuine usefulness.
It's a single practitioner study on one engine, so read it as directional. But it echoes a pattern other 2026 analyses report and that we've covered in why AI loves listicles: for commercial-intent questions, third-party roundups beat brand-owned pages by a wide margin.
Why engines reach for roundups over your own page
When a shopper asks "what's the best X," the model wants a neutral-looking, multi-option answer. A page that already weighs several brands against criteria matches that shape; a page listing only your own products reads as a sales pitch and gets discounted.
You can't out-neutral an independent comparison by writing your own "best tools (and we're #1)" page. The model is trained to treat single-brand roundups as biased, so the work isn't more self-promotion, it's getting into the lists that already get cited.
Roundups also carry higher entity density. Many named brands co-occurring with the category, and more verifiable specifics, both of which retrieval favours.
The earned-placement playbook
If the lists win and you can't credibly publish your own, the strategy is to earn your way into the ones that already get cited:
- Find the lists AI is actually pulling. For your top buyer questions, see which roundups the engines cite, and treat that set as your target list, not a random outreach spread.
- Earn placement, fairly. Pitch reviewers and publishers with verifiable specifics. Pricing, features, real differentiators. The things a list editor needs to slot you in accurately.
- Aim for the top third. Position inside the list matters; a mention buried at the bottom does less than a spot near the top. Give editors the evidence to rank you well.
- Prioritise freshness. Lists updated in the last quarter get cited most. Favour publishers who keep their roundups current, and revisit your own placements so your listed facts stay accurate.
- Mind authority. Both ways. Plenty of cited lists sit on low-authority domains, but durable visibility comes from credible ones. Chase placements that are both cited and trustworthy.
Where your owned content still wins
This isn't an argument against publishing. It's about giving owned content the right job. Your own pages win brand- and feature-specific questions: "how does X work," "X pricing," "does X integrate with Y." There the model wants the canonical first-party source, and that's you. Reserve owned content for those queries plus genuine educational explainers that build entity strength; pursue earned placement for the "best X" questions.
The net: for "which is best" buyer questions, the independent roundup wins, so the move is getting into the comparison content that already gets cited, and then tracking whether it's working across every engine. That tracking is what Buffy Intel is for.