In short: Roughly 44% of the pages ChatGPT cites for “best X” questions are listicles, and many of them are written by the brand that ranks itself first. That is the AI listicle loophole. New data from mid-2026 shows Google now cites those self-promotional lists while recommending the competitors named inside them 69% of the time. The shortcut is closing, and the content that keeps winning is the content built to be useful.
Have you noticed that whenever you search for “the best” or “highest-rated” product or service in ChatGPT, Gemini, Claude, or any other AI tool, you get a strangely similar list?
Even weirder is that many of the cited sites have nothing to do with the actual product or service. They are content farms presenting a listicle of companies in more or less random order.
This is the AI listicle loophole, and it is quietly degrading the quality of AI search results.
What Is the AI Listicle Loophole?
The AI listicle loophole is the practice of publishing a “best [category]” list on your own site, ranking yourself first, and letting AI search engines cite it as if it were an independent recommendation.
An Ahrefs study of 26,283 source URLs found that roughly 44% of the pages ChatGPT cites for top-of-funnel “best X” prompts are blog listicles. A large share of those are written by the same brands that put themselves in the top spot.
It is the easiest, cheapest, least defensible way to show up in an AI answer, and every agency with a content calendar and a ChatGPT or Claude subscription is already running the play.
So let’s talk about whether you should run it too.
What Changed Since April 2026
When I first wrote this post in April, the AI listicle loophole looked like a free lunch. Five months later, the bill has arrived.
In June, Lily Ray published an analysis of 100 B2B “best software” queries in Google AI Overviews, tracked across three dates from April to June. Self-promotional listicles were cited 323 times. In 224 of those cases, Google cited the brand’s own page and then recommended someone else.
Read that again. 69% of the time, the brand did the work of writing the list, Google used it as a source, and the recommendation went to a competitor named further down the page. One learning-platform company wrote a “best LMS” list, ranked itself first, and watched Google recommend Kajabi, Thinkific, LearnWorlds, and Teachable instead.
A citation is not a recommendation. That distinction is the whole game now.
Ray also tracked organic traffic for dozens of sites that leaned heavily on self-ranked “best” pages. Declines started around January 20 and accelerated through Google’s May 2026 core update. Search Engine Land separately reported that the tactic may carry legal risk under the FTC’s Consumer Review Rule when company-controlled content is presented as an independent review.
And honestly, the Ahrefs data already hinted at this. Their study found that 79% of the “best” lists ChatGPT cited had been updated within the year, and 26% within the previous two months. AI engines reward freshness. A listicle you published once and forgot is exactly the page that stops getting cited first.
What the Data Actually Says
Google’s AI Overviews, ChatGPT, Perplexity, and the rest of the answer engines pull from the live web every time they answer a question. They don’t wait for a new training run. What’s indexed today becomes tomorrow’s citation.
That means the content winning in AI answers right now is whatever’s most quotable. Accuracy and usefulness come second. Listicles are quotable by design. They’re structured, enumerated, and easy for a language model to extract. A “best CRM for small business” post with ten bolded headings and a summary table is a buffet for a retrieval engine.
Hence the 44%. And hence the self-ranking problem.
How the Self-Ranking Trick Works
Here’s the play. A software company writes a post titled “The 7 Best [Category] Tools for 2026.” They rank themselves first. They include six competitors, described accurately but with softer language. They publish on their own domain, which has decent authority. ChatGPT crawls the web, sees a structured listicle from a plausible source, and cites it when someone asks “what’s the best [category] tool for small business?”
The user sees a confident answer with a citation. The citation goes to a page where the brand that wrote it ranks number one. The user clicks. The brand wins.
Twenty minutes of work. Real pipeline impact. No outright lies told. Just a structured suggestion that the model is happy to repeat.
Or at least, that is how it worked in April. The June data says the model is now happy to repeat the list and skip the author.
Why AI Listicles Are Still Working (for Now)
There are three reasons the AI listicle loophole is still producing citations, even as the recommendations drift elsewhere.
First, AI engines are hungry. They have to cite something, and a clean listicle from a B2B SaaS blog looks far more structured and authoritative than a Reddit thread where the real answer often lives. The model picks the form that’s easier to parse.
Second, most users don’t check. The whole promise of AI search is that you don’t have to verify. You ask, you get an answer, you move on. That’s the product.
Third, the free tier makes it worse. Search Engine Journal reported that paid AI tools are 26-33% better at avoiding hallucinations and manipulation, while the large majority of ChatGPT’s weekly users are on the free tier. Your prospects are overwhelmingly getting the version of AI search that’s easiest to game.
So yes, the AI listicle loophole still gets you cited. The real question is whether a citation without a recommendation is worth anything, and the answer is increasingly no.
The Cost of Copying the Trend
The obvious issue is that the whole industry is running the same play. If every SaaS company writes a “7 best tools” post and ranks itself first, the signal gets noisy fast.
What wins, then, is whoever has the biggest content pipeline, churning out the most listicles. That’s a race to the bottom, and it’s one where small businesses and solopreneurs can’t outspend the content farms. If the game is volume, you lose. If the game is gaming the game, you also lose.
There’s also a trust problem, and it is no longer coming. It is here. Google is already discounting the self-ranked brand while keeping the list. Users are starting to notice too. When they discount listicle citations the way they discounted banner ads, the brands that built their whole AI visibility strategy on self-ranking listicles have nothing else in the tank.
We wrote about something similar in AI Marketing Trends 2026: Why Helping People Decide Still Wins. The tactics that look clever in 2026 tend to age like milk. The tactics that compound are the ones built around actually helping someone make a decision.
What Can You Do Instead?
The better play is the one most people won’t run because it takes longer. Build content that AI engines cite because it’s the most useful answer on the web. Structure helps, but usefulness is what earns the recommendation.
Three practical shifts for small businesses and consultants:
Write for the question. AI answer engines are trained to match intent. A post that thoroughly answers “how do I choose a CRM if I’m a solo consultant with 30 clients?” will outperform a generic “best CRM” listicle over time, because the specificity matches how people actually ask AI. For a longer breakdown of this idea, see our complete guide to AI marketing for small business.
Cite your own data. Original research, proprietary frameworks, and firsthand case studies are the hardest content to manufacture at scale. They’re also the content AI engines prefer to cite, because there’s no upstream source to link to instead. You become the source other listicles have to reference. If you want to see what a list looks like when it is built from actual reading rather than a keyword tool, our AI marketing books post is the model we use.
Be visible in the places AI pulls from. Reddit, YouTube, and trade publications were among the most-cited domains in Ray’s “best” query data, and Reddit citations rose sharply through the spring. If you’re a consultant, a regular presence in the right subreddit can be worth more than another listicle post on your blog.
Underneath all three shifts is the same idea we covered in Why Your AI Marketing Strategy Still Needs a Human Touch. Machines can copy structure. They can’t manufacture lived experience or a real point of view. The content that survives the retrieval wars is the content that has something a listicle doesn’t. That applies to how you use AI in the writing itself, too, which is why we are careful about where AI blog writing helps and where it hurts.
AI Listicle Loophole FAQ
Should I write “best of” listicles for AI search?
Only if they are honest and useful. A list that discloses how it was put together, links out to the alternatives, and gets updated regularly can still earn citations. A list that exists to rank yourself first is now more likely to send the recommendation to a competitor than to you.
How does ChatGPT decide which pages to cite?
ChatGPT retrieves from the live web and favors pages that are structured, recently updated, and easy to extract an answer from. The Ahrefs data shows listicles, comparison sites like G2, and landing pages dominate top-of-funnel citations, with freshness mattering more than it does in traditional Google rankings.
Is being cited by AI the same as being recommended?
No. A citation means the engine used your page as a source. A recommendation means it told the user to pick you. The mid-2026 data shows those two things separating fast, especially for self-promotional lists.
What should a small business do instead of the AI listicle loophole?
Answer specific questions your customers actually ask, publish something original that other sites have to reference, and show up in the third-party places AI engines already trust. Slower, less clever, and it holds up.
Loopholes Disappear, Quality Does Not
The AI listicle loophole is a symptom of a bigger shift. The SEO industry spent twenty years learning to game Google. A chunk of that industry is now applying the same playbook to AI search, and it worked well enough for a while that the shortcuts felt tempting.
But the cost of playing is that you build a business that only works while the loophole is open. The retrieval layer is getting smarter, the users are getting more skeptical, and the content that wins is the content that was built to be useful from the start.
That’s the strategy we’re backing at Shorthand. It’s slower, it’s less clever, and it’s aging better already.
If you want help building an AI marketing strategy that doesn’t depend on a shortcut, get in touch. We’ll take the long road with you.
– Daniel Midson-Short