How to Publish a 'Best X' List AI Will Cite: 12 Checks

What Ranqo's Research Found

The paper measured ChatGPT, Claude, Gemini, Grok and Perplexity. Four things it reports matter if you're publishing a list.

Ranked lists took the largest share of content citations

When an engine cited a page, about 59% of the time it was content, such as an article, guide or video, rather than a homepage or product page. Ranked lists took 35.7% of those citations, ahead of generic articles at 31% and how-to guides at 9.7%. Pages built as a direct comparison took 4.5%. Our post on the page types AI cites most goes through the top five.

One list can be reused across many questions

The paper calls the ranked list the highest-leverage page a brand can target: once a list includes a brand, that one page becomes a source the engines reuse across many different questions. These lists appeared on media sites, blogs and review sites alike.

Few citations went to a brand's own site

Only 2.9% of citations went to a brand's own domain, while 75.2% went to other companies in the same space, including their product pages and corporate blogs. So engines do cite company pages, but your own list will be competing with all of theirs.

Big brands were named anyway, and specific questions were harder

The first time a brand was measured, on questions that didn't mention it, global household names appeared in an average of 73% of answers, mid-market brands in 44% and niche brands in 11%. Separately, on broad questions such as “best running shoes”, a given brand appeared in 22.9% of answers on average, against 10.8% on questions about a specific problem and 11.4% on a specific use case. A brand can be findable for the broad ask and nearly invisible for the specific one.

These figures describe March to May 2026, on the models those engines ran then. The paper only observed what engines cited, so it can't show that publishing a list would earn you more citations. The checks below are what a list should do either way.

The 12 Checks

Google treats a ranked list as a review: its reviews system covers “ranked-lists of recommendations”. So what Google asks of a review applies to your list.

1. Put the picks at the top

Name every pick in a short summary or table before the long reviews start, so a reader gets the answer first. Ranqo's research paper gives the same advice for any page: state your key facts clearly and early, where the engine is most likely to read them.

2. Rank the real rivals, including any that beat you

A list that leaves out the obvious options, or puts your product first without a reason, reads as an advert. Include the products buyers actually compare you with and rank them on your method. Leaving a big name off probably won't keep it out of the answer: in the paper, household names appeared in an average of 73% of answers on their first tracking run, to questions that didn't mention them. If you are the best fit for one kind of buyer, say so in that pick's label rather than taking #1 for everyone.

3. Say how you chose

Say what you tested, how, and how many products you looked at. Google suggests building trust by showing “the number of products that were tested, what the test results were, and how the tests were conducted”.

4. Back every pick with evidence

Google's review guidance says that when you recommend something as the best, you should include why you consider it the best, with first-hand supporting evidence, and share quantitative measurements of how it performs. Screenshots, test results, measured numbers and your own photos do this; a paragraph rewritten from the vendor's homepage doesn't.

Citing your sources may help too. In a simulated AI search engine, a 2024 study found that adding quotations, statistics or source citations each raised a page's visibility by 30 to 40% on its main measure. The study's prompts let the model make up what it added, so add only ones you can link to.

5. Say who each pick is best for

A “best for” label gives one ranking several possible answers: best for small teams, best on a budget, best for beginners. Google asks reviewers to explain which might be best for certain uses or circumstances. The label also speaks to the questions where brands were hardest to find: in the paper, a given brand appeared in 11.4% of answers to use-case questions on average, about half the 22.9% on broad questions.

6. Add a comparison table

Set the picks side by side in one table, on the same measures for each, such as price, who it suits and its main limit. A table is one way to do what Google asks of a review: explain what sets something apart from its competitors.

7. List prices, with the date you checked them

A buyer who has a shortlist asks what each pick costs. Put the month you checked beside the price: prices change more often than the rest of the page, and a stale one is an error any reader can spot. If a pick has no public price, say so.

8. Give pros and cons, your own product’s included

Google's review guidance asks you to discuss the benefits and drawbacks of what you review. Pros and cons show you used the products, and on a vendor's list, a real drawback of your own product is good evidence that the rest of the list is fair.

9. Disclose your stake

If you make one of the products, or earn a commission on any of them, say so in one plain line above the first pick, for example: “We make Acme, #3 below, and tested it the same way as the others.” A reader who finds out later has reason to doubt the rest of the page, while a vendor's list that says so plainly can still be the most useful list on the subject. Being a company doesn't rule a page out: in the paper, 75.2% of citations went to pages from other companies in the same space.

10. Update for real, and say so on the page

Update when the picks, prices or tests change, and show the date where readers see it. Don't change the date, or the year in the title, on its own: Google asks whether you are changing the date of pages to make them seem fresh when the content has not substantially changed.

11. Cut any pick you can’t back

Stop at the picks you can back. A longer list makes it harder to give every pick the evidence, the pros and cons and the dated price the other checks ask for, and Google asks that a ranked list carry enough useful content to stand on its own. For most categories, we'd expect that to leave about ten.

12. Make sure crawlers can read it and Google can show it

A list that only appears after a script runs, or sits on a server that turns away crawlers it doesn't recognise, is hidden from any engine that can't get past that. To check yours, run the URL through Ranqo's free AI Crawler Inspector, which fetches it as each AI crawler and as a browser and flags any that get turned away or a page with little text.

For Google's AI features, the page must be indexed and allowed a snippet. Google lists the snippet controls that limit what its AI features show, so check that no nosnippet, data-nosnippet or restrictive max-snippet rule covers the list. Then check that the site's Search generative AI control in Search Console still includes it.

What to Avoid

• Lists written to steer AI answers. Google's spam policies count attempting to manipulate generative AI responses in Google Search as spam, a clarification Google published on 15 May 2026.
• A list for every keyword. Dozens of near-identical “best X for Y” pages made to rank rather than to help can fall under scaled content abuse in the same policies.
• A list anyone could write. Google's guide to its generative AI features uses a numbered listicle, “7 Tips for First-Time Homebuyers”, as its example of commodity content, which it says “typically adds little unique insight for readers”. A “best X” list of the obvious picks, with nothing tested, has the same problem.
• Relying on schema. Google says there is no special schema.org markup to add for its generative AI features, and Ranqo's research paper treats schema as hygiene rather than a citation lever, a view it has yet to test. Our guide to schema markup for AI citations covers what markup does do.

For the rest of the page, see our checklist for getting cited in AI search.

After You Publish

Run check 12 on the live URL, then ask each engine the questions your list answers and see whether it cites the page. Ranqo's research paper advises treating each engine as a separate market: it says the sources they cite overlap little and that a win on Perplexity doesn't transfer to Gemini. It also advises checking again 14 to 30 days after a change and adjusting from there.

How Ranqo Fits

Ranqo shows which lists AI cites for your own buyer questions. On the Sources page's URLs tab, every cited page carries its content type, so picking Ranked List from All Content Types shows the lists the engines read for your category and the share of answers that cited each. Once yours is live, you can see there whether answers cite it. Outreach turns the lists on other sites into a worklist: its Publications channel ranks the cited lists, comparisons, reviews and alternatives pages you could pitch for a place on, leaving out your own site and your competitors', and drafts a pitch for you to edit and send yourself.