Obility Editorial · · 4 min read
Self-ranked lists do show up in AI answers
Search for the best software in almost any B2B category and you will find vendors ranking themselves first on their own blogs. The tactic spread because it appeared to work in AI answers, and the best public evidence says it did.
Ahrefs published a study in December 2025 in which Glen Allsopp reviewed ChatGPT answers to 750 top-of-funnel prompts about software, products and agencies, and categorised the sources behind them. “Best X” blog lists made up 43.8 percent of all source page types. In the software prompts, when ChatGPT recommended a brand, that brand’s own self-promotional list appeared among the sources in more than a third of responses. In Google’s results, 169 of 250 “best X software” searches showed a list in which the publisher ranked itself first.
Two things in the same study deserve as much attention. About 35 percent of the cited lists sat on low-authority domains, many of which Allsopp calls highly questionable, and he expects that share to fall as assistants get better at judging quality. And the data shows what gets cited, not what causes a recommendation. For a “best X” prompt, lists are cited because lists are most of what exists to cite.
The citation rests on a search ranking that can move
An answer engine that searches the web before answering can only cite pages its search step finds. Google’s AI Overviews and AI Mode draw on Google’s index, and other assistants rely on search results of their own. A self-ranked list earns its AI citations by ranking, so it is exactly as durable as that ranking.
In February 2026, the SEO consultant Lily Ray published an analysis of several SaaS and B2B sites that lost between 29 and 49 percent of their Google visibility from mid-January, as measured by Sistrix. Their blogs drove most of the losses, and each carried self-promotional “best” lists, from 10 on one site to 340 on another. Ray suspects refinements to Google’s reviews system. She also reports that Glenn Gabe saw four of the same sites losing visibility in AI Overviews.
Read the evidence with its limits. Ray notes that these sites also scaled content quickly, refreshed dates without real updates and misused review markup, and that the lists were a small share of their pages: 191 out of roughly 30,000 on the largest blog. Grow and Convert, a content agency, replied that its own clients’ comparison posts still perform and that Google is penalising bad content, not the format. Nobody has published traffic data isolating these pages, so a crackdown is plausible, not proven. What nobody disputes is that the citation depends on the ranking.
Google has already described the list that holds up
Google’s documentation on its reviews system says it evaluates first-party content written to recommend, including head-to-head comparisons and ranked lists, and that a site with a lot of review content can be assessed as a whole. Its guidance on writing high quality reviews asks for things a vendor-written list rarely has: benefits and drawbacks based on original research, coverage of comparable options and which suits which situation, links to other resources, and, when something is called the best, the reason why with first-hand supporting evidence.
A vendor that ranks itself first among competitors it has never used cannot meet that last test. It can still publish a useful comparison. State plainly that you are the vendor. Set out the criteria a buyer should use. Describe each alternative fairly, with the kind of team it suits, and link to it. Then place your own product where it honestly fits, by situation rather than by rank. Ahrefs, after publishing its study, said it would keep such posts under half a percent of its blog, would not always put itself on top, and would keep linking to alternatives. Grow and Convert says it avoids forced rankings altogether.
Other publishers’ lists carry more of the weight
The same Ahrefs study found a correlation between ranking highly on third-party lists and appearing in ChatGPT’s recommendations. It also found that, in the software category, a recommended brand’s own landing pages appeared among the sources slightly more often than its own list did, 37.2 percent of responses against 34 percent. So the pages that describe what you actually do, and the lists other people publish, matter at least as much as the list you write about yourself.
That makes the useful audit a question of sources, not formats. For each buying prompt you care about, record which lists the answer cites, where you appear on each, and whether your own page is among them. In Obility, Answer Engine Insights reviews the surrounding answer, the competitors it recommends and the sources it cites for each prompt, and gives every gap an owner. As an illustration, if the same three industry lists keep being cited across your category prompts and you are missing from two of them, the work is outreach and positioning with those publishers, not another post on your own blog.
One honest comparison beats a hundred rankings
Do not read the 43.8 percent figure as a recipe. It describes what existed to cite in late 2025, before the January volatility Ray documented, and the sites that lost visibility had produced lists by the dozen. Write one honest comparison for each category you genuinely compete in, keep it current when the market changes, and stop there.
Then measure it on two levels. Watch the page’s organic ranking, because a fall there arrives before a fall in the answers that cite it. And track the prompts it was written for, so you can see whether the answers cite your comparison, your landing pages, or a third-party list that names you. If only your own list ever mentions you, the answers are telling you that nobody else does yet.