Should you write listicles about your business?

Why are businesses writing “best tools” listicles about themselves?

Businesses are writing self-promotional “best tools” listicles because these pages have been unusually visible in both Google and AI answers, and for a while the tactic was almost ridiculously easy to exploit.

Ahrefs analyzed 26,283 source URLs used for recommendation-style AI queries and found “best X” lists appearing in 43.8% of ChatGPT source sets. Wix Studio later analyzed more than one million citations from 75,000 answers across ChatGPT, Google AI Mode and Perplexity. Listicles accounted for roughly 40% of citations when the prompt had commercial intent.

Companies noticed. In Ahrefs' review of 250 Google results for “best X software” searches, 67.6% contained a list where the company publishing the article ranked its own product number one.

The logic is easy to understand. A CRM company can publish “10 Best CRM Tools,” put itself first, describe nine competitors underneath, and create a page that closely matches what someone might ask Google or ChatGPT.

That worked well enough to become a real SEO and GEO tactic. The harder question now is whether copying it is still smart.

Do LLMs really cite listicles more than other content?

Yes, LLMs still cite listicles heavily for commercial questions, although the newest data suggests the format has lost some of its earlier advantage.

The Wix Studio study is one of the clearest large datasets here. It examined 1,056,727 citations across ChatGPT, Google AI Mode and Perplexity. About 40% of commercial-intent answers cited listicles, almost twice the rate seen for other intents.

DeltaV Digital found an even bigger effect in one specific market. In its analysis of more than 25,000 AI citations across eight industries, listicles generated 61% of citations for B2B technology services.

But the same DeltaV research found zero listicle citations in healthcare, where articles and authoritative medical sources were far more common.

The pattern is fairly clear. When someone asks for choices, alternatives or the “best” products, a page that already compares several options is convenient source material. When someone asks for factual or technical information, other formats often work better.

So listicles are still unusually relevant for commercial searches these days. Treating them as a universal trick for AI visibility would go much further than the evidence supports.

Research	What it found
Wix Studio, 1.05M AI citations	\~40% of commercial-intent answers cited listicles
DeltaV Digital, B2B technology services	61% of citations came from listicles
DeltaV Digital, healthcare	0% came from listicles

How often do LLMs cite self-promotional listicles?

Self-promotional listicles currently get through AI retrieval systems often enough to work, but they represent only a minority of the listicles LLMs cite.

Peec AI tracked 232,000 citations across 13,000 listicles over 12 weeks and specifically looked for pages where companies recommended their own products.

Roughly 11% of citations came from those self-promotional pages.

The differences between AI systems were large. ChatGPT had the lowest rate at around 4%, while Google AI Mode and Perplexity were closer to 10% to 11%.

Peec also looked for a steady decline during its observation period and could not find one. In other words, the AI platforms had not yet developed a reliable filter that simply excluded a page because the company writing it also appeared in the ranking.

That makes the loophole real.

It also puts the bigger listicle numbers into perspective. When we hear that 40% of commercial AI answers cite listicles, that does not mean 40% are citing companies declaring themselves the best. Most of the citation pool comes from other kinds of lists.

AI citation finding	Approximate share
All self-promotional listicles in Peec sample	11%
ChatGPT self-promotional listicles	4%
Google AI Mode / Perplexity	10–11%

If your listicle gets cited, will the AI recommend your business?

Often no. Google can cite a company's own “best software” list and then recommend competitors taken from that exact page.

Lily Ray tested 100 B2B “best [category] software” searches that produced 80 Google AI Overviews. Self-promotional listicles appeared as citations 323 times.

In 224 of those cases, the company publishing the listicle was missing from Google's recommendations.

That is 69%.

Imagine that we run a project-management tool and publish “10 Best Project Management Tools.” Google may use our comparison to understand Asana, Monday, ClickUp and our own product, then recommend Asana and Monday while citing our article as supporting evidence.

Ahrefs later reproduced the same basic problem in an experiment using its own promotional content. When an AI answer cited one of Ahrefs' pages promoting its Evolve conference, 43% of those answers still failed to mention Ahrefs Evolve. When the page had been retrieved behind the scenes without receiving a visible citation, the conference disappeared from 74% of answers.

So citation counts can be dangerously flattering.

If our goal is revenue, the useful metric is whether the AI actually names or recommends our company. A page that gets cited while feeding competitors into the answer may be doing excellent GEO work for everybody except us.

Does ranking your own product #1 actually help?

We currently have very little evidence that putting your own product first causes a listicle to perform better.

The habit is certainly common. In research published just days ago, Kevin Indig manually reviewed vendor-owned B2B listicles and found that 74.7% of eligible comparisons put the publisher's own product first.

At first glance, those self-promotional pages looked extremely strong. Their measured traffic had grown far faster than the broader listicle sample.

Then the analysis controlled for starting traffic, industry, listicle type, current ranking position and repeated pages from the same publisher.

The advantage almost disappeared.

Putting your own product first was associated with only 4% stronger growth. The 95% confidence interval stretched from a 51% decline to a 120% increase, and the p-value was 0.91. Statistically, there is no useful evidence there that self-ranking caused the growth.

The more plausible explanation is that companies investing seriously in these pages also tend to produce better pages, update them more often and target valuable commercial searches.

We can still rank ourselves first when the comparison genuinely supports it. Doing it automatically because someone said LLMs reward position one is much harder to justify.

Do self-promotional listicles work better for unknown businesses?

Yes, self-promotional content seems most useful when an AI simply does not know enough about a legitimate smaller brand yet.

Ahrefs' experiment produced a surprisingly clean example.

Ahrefs Evolve was a relatively new conference. The researchers first found prompt-and-engine combinations where the conference never appeared. After Ahrefs published promotional pages, Evolve entered 72 previously empty combinations.

In 82% of those new appearances, one of the new promotional pages was cited.

The result for the established Ahrefs brand was completely different. When Ahrefs appeared in previously empty answers about its Brand Radar product, only 6% of the new appearances coincided with the experimental pages. Third-party sources explained almost everything else.

Relevance also changed the effect dramatically. Evolve appeared in 66.4% of tracked answers about “best SEO conferences,” where it clearly belonged, compared with only 15.8% for the broader “best marketing conferences.”

The boundary is pretty clear.

A good first-party comparison can help an AI learn that a lesser-known company genuinely belongs in a category. Trying to force a weak category association is much less convincing because the rest of the web is still telling the model something different.

Once ChatGPT starts citing your listicle, does the citation stick?

No. ChatGPT and other AI citations remain unstable, so one good screenshot tells us very little about long-term visibility.

Ahrefs tracked its promotional pages across 9,886 AI answers and found that roughly a quarter of page-query-engine combinations were cited once and never again.

Even between the first and last citation, a page appeared on only about one third of the days when it could have appeared.

Ahrefs has seen similar volatility at much larger scale in Google's AI results. In an analysis covering more than 43,000 AI Overview keywords, only about 54% of cited URLs remained the same from one check to the next.

That is very different from the old mental model of reaching position one in Google and staying there for weeks.

AI systems rerun searches, rewrite fan-out queries, pull different documents and change their answers after model updates. A listicle can influence ChatGPT today and disappear from the evidence set soon afterward without anything on the page changing.

So we should measure repeated recommendation visibility over time. A single citation is closer to one successful retrieval than a permanent ranking.

Is ChatGPT already moving away from listicles?

Yes. ChatGPT's latest retrieval behavior has reduced listicles' share of citations sharply, and this is probably the strongest warning against building a whole strategy around the format.

Peec AI compared around 180 million sources across one million tracked prompts before and after the GPT-5.6 change.

Listicles fell from 15.77% of classified citations to 7.80%, a relative drop of 50.5%. Comparison pages fell from 9.08% to 6.17%, down 32.1%.

The searches ChatGPT runs behind the scenes also changed. Terms such as “best,” “top,” “reviews,” “comparison” and “vs” appeared less often, while searches targeting specific sites, official sources and pricing information became more common.

There is an important wrinkle, though. ChatGPT also started consulting far more sources per answer. Other researchers measuring the same transition found the source pool roughly doubling.

That means the absolute number of listicles encountered per answer may have fallen much less than the 50.5% headline suggests. Their share of ChatGPT's research mix is what clearly collapsed.

We cannot prove that OpenAI deliberately targeted GEO listicles. A model update changes many things at once.

But we can be much firmer about the practical conclusion: a format marketers considered one of the safest AI-citation hacks has already lost a large part of its relative advantage after a single model change.

ChatGPT source type	Before GPT-5.6	After GPT-5.6	Change in share
Listicles	15.77%	7.80%	-50.5%
Comparison pages	9.08%	6.17%	-32.1%

Is Google Search demoting listicles now?

No, Google still shows a huge number of listicles. What has changed lately is which listicles win.

A fresh analysis of 5.3 million Google results across almost 60,000 US searches found at least one genuine listicle in the top ten for 55.1% of queries. A listicle appeared in the top three for 32.3%.

Search intent made a huge difference. Google showed 5.3 times more listicles when the query explicitly asked for a set of options.

The rankings were volatile, though. Only 52.5% of queries kept the same number-one listicle URL between the two periods studied.

Publishers also had an obvious edge over brands. Publishers supplied 46.8% of the highest-ranking listicles, while brands and vendors supplied 19.2%. When a publisher listicle and brand listicle appeared against each other in the same results, the publisher won 54% of those direct matchups.

So Google has not decided that listicles are bad content. People genuinely want lists when they search for products, ideas and alternatives.

The weaker bet today is assuming that a vendor-owned “we ranked ourselves first” page deserves the same trust as an independent comparison simply because both use the same format.

Are self-promotional listicles black-hat SEO?

No. Writing an honest comparison that includes your own product is allowed by Google's published rules.

Google's own review guidance explicitly says reviews can come from an expert staff member or a merchant helping people choose between competing products. Google's reviews system also covers head-to-head comparisons and ranked recommendation lists.

The quality requirements are fairly clear. Google wants original research, evidence of real experience, quantitative measurements, meaningful differences between products, advantages and disadvantages, and an explanation of why something deserves a recommendation.

The risk starts when companies take the listicle insight and scale it into low-value search inventory.

Google defines scaled content abuse as creating many pages mainly to manipulate rankings rather than help users. Its examples include mass-generating pages with AI, stitching together existing information without adding value and producing lots of keyword pages that offer little useful information.

One serious “Best CRM Tools for Freelance Recruiters” comparison where we tested every product can fit comfortably inside Google's review guidance.

Publishing hundreds of almost identical “Best CRM for Dentists,” “Best CRM for Plumbers,” “Best CRM for Architects” and “Best CRM for Accountants” pages because an SEO tool found the keywords is much harder to defend.

Calling the whole strategy black hat is too simplistic. The danger comes from how aggressively we manufacture pages and how little original value those pages contain.

Can a self-promotional listicle hurt people's trust in your business?

Yes. Ranking yourself number one while pretending to be an independent reviewer can make the whole comparison look rigged.

We do not have a giant randomized study specifically measuring trust in SaaS companies that write “10 Best SaaS Tools” pages about themselves. The adjacent advertising research is still useful, though.

Experiments on sponsored comparative recommendations have repeatedly found that perceived commercial manipulation lowers trust and purchase intent. Research comparing company-owned content with third-party review environments also tends to find stronger persuasion when the recommendation comes from an independent source.

There is a simple human explanation. Readers understand that HubSpot has an incentive to rank HubSpot highly in “Best CRM Software.” The article becomes suspicious when it writes about HubSpot as if some detached editorial board happened to discover the company and crown it the winner.

Disclosure makes that easier to swallow.

Something like “We make one of the products below, so we used the same five criteria for every tool and included the raw results” immediately gives the reader a way to judge the comparison.

We lose some fake objectivity but gain credibility.

For a business we expect to operate for years, that is an easy trade.

Could self-promotional listicles create legal problems?

Yes, but the legal risk comes mainly from misleading comparisons and unsupported claims rather than from mentioning competitors.

The US Federal Trade Commission has long allowed comparative advertising when the comparisons are truthful, clear and supported by evidence.

So a company can say that its software costs less than Salesforce or processed a test faster than HubSpot if the comparison is accurate and reasonably substantiated.

Problems appear when apparently objective claims cannot be proved.

“Fastest CRM,” “best customer support,” “highest-rated platform” and “50% cheaper than competitors” can sound like harmless listicle language, but they may also be factual advertising claims.

The safer version is specific enough that somebody else could check it. We might say that our product ranked first in our test because it was the only one among the ten products examined that offered three named features below a stated price.

That is still promotional. At least the reader can see exactly how we reached the result.

Is LLM traffic big enough to justify changing your whole content strategy?

No. AI referrals are growing really fast, but current traffic data still makes it hard to justify wrecking a good SEO or brand strategy just to chase ChatGPT citations.

A peer-reviewed Marketing Science study analyzed first-party analytics from 973 ecommerce websites with $20 billion in combined revenue and more than 50,000 ChatGPT-referred purchases.

LLM referrals represented less than 0.2% of total traffic in that dataset, roughly 200 times less than Google organic search. ChatGPT visitors converted better than paid social but worse than every other traditional channel studied.

More recent traffic research shows how quickly things are changing. Previsible analyzed 6.77 million LLM-driven sessions and found monthly AI referrals had increased 9.9 times from its earlier baseline. ChatGPT generated 92.4% of those visits.

Both findings can be true at once.

AI discovery is growing from a small base, and it can already matter a lot for particular businesses. Yet the average company still gets much more traffic and revenue from established channels.

A durable comparison page still makes sense. Turning an entire website into an AI-citation farm looks much harder to justify.

Is getting into other people's listicles better than writing your own?

Usually yes. Independent mentions currently look like a stronger long-term asset than publishing more pages where your company recommends itself.

Ahrefs studied 75,000 brands and compared dozens of factors with visibility in ChatGPT, Google AI Mode and AI Overviews.

Branded web mentions showed correlations of roughly 0.66 to 0.71 with AI visibility. YouTube mentions were even higher at around 0.74.

The number of pages on a company's own website had almost no relationship with AI visibility by comparison.

We have to be careful here because correlation does not prove that outside mentions directly cause AI recommendations. Bigger brands naturally receive more mentions and more AI visibility.

Still, the gap is large enough to influence strategy. AI systems repeatedly encounter a brand across publishers, YouTube videos, specialist websites, customer discussions and comparison pages. That gives them much more independent evidence that the company belongs in a category.

As seen above, publisher listicles also beat brand-owned listicles in the newest Google ranking research.

If we have ten hours available, spending all ten writing another ten articles about ourselves would be a strange allocation. Getting one genuinely useful comparison onto our site and then earning independent mentions elsewhere gives us a much stronger information footprint.

What kind of self-promotional listicle is safest long term?

The safest self-promotional listicle is one we would still want on our website if ChatGPT stopped citing listicles tomorrow.

That test clears away most of the questionable versions immediately.

A useful comparison solves a real buying problem. “Best Accounting Software for Three-Person Consulting Firms” gives us a much clearer job than “10 Best Accounting Platforms” because we can test the things that actually matter to a three-person consulting firm.

The methodology should also be visible. We should explain which products we tested, which prices we checked, which features mattered, what competitors do better and which customers should choose something else.

Google's own review guidance asks for essentially the same things: first-hand evidence, quantitative measurements, meaningful differences, advantages, disadvantages and a reason for the final recommendation.

Freshness matters too. In the newest Google listicle research, 66.6% of top-three listicles displayed a date from the previous two years, compared with 57.3% of pages ranking eighth through tenth. The study cannot prove that updating the date itself improves rankings, but stale comparison content clearly appears less often among the strongest survivors.

That creates a cost people underestimate.

Every listicle we publish becomes something we have to maintain. Prices change. Features disappear. Free plans get killed. Competitors improve.

If we have no intention of checking the comparison again, we probably should not publish it in the first place.

Should you publish dozens of “best X” listicles about your company?

No. Publishing dozens or hundreds of self-promotional listicles is currently the weakest version of this strategy.

The first problem is diminishing informational value. Once we have a strong “Best Project Management Tools” comparison, another 80 pages changing only the profession, team size or adjective rarely add 80 times more useful information.

The second problem is Google's scaled-content-abuse policy. Volume alone does not violate the rule, but producing large numbers of low-value pages mainly to manipulate search rankings sits very close to Google's published definition.

The third problem is AI retrieval itself. As pointed out above, ChatGPT has already sharply reduced listicles' share of its citation mix. A company that produced 500 pages to exploit one retrieval behavior has far more downside from the next model update than a company that produced five pages because customers genuinely needed them.

There is even a useful clue from Ahrefs, one of the companies publishing some of the strongest evidence that self-promotional content can work. Ahrefs has said comparison-style content remains a tiny fraction of its overall blog.

That feels much closer to the right ratio.

A few strategic comparison pages can cover important buying decisions. Turning the entire editorial strategy into fake Wirecutter is asking an algorithm to eventually notice what human readers already can.

What should you do if you want ChatGPT and other LLMs to recommend your business?

We should give LLMs solid information about our business on our own website, then work much harder on getting the same story confirmed elsewhere.

Our website still needs clean product pages, pricing, documentation, use cases, comparisons, original data and clear explanations of who the product is for.

But publishing more first-party pages alone looks like a weak way to build recommendation authority.

The 75,000-brand Ahrefs study found much stronger relationships between AI visibility and broad web mentions than between AI visibility and sheer page count. Independent listicles, specialist publications, YouTube, communities, customer reviews and other third-party sources all help create repeated associations between a brand and a category.

Original research can be particularly effective here.

If we survey 1,000 marketers and publish a genuinely useful benchmark, the report lives on our site first. Journalists may cite it. Bloggers may use the numbers. YouTubers may discuss it. Other listicles may reference the findings. One original dataset can spread our name across dozens of independent sources.

Compare that with producing twenty articles whose core message is “our product is great.”

The first approach creates evidence other people want to reuse. The second creates twenty versions of our own opinion.

Should you write listicles about your business?

Yes, but we would publish a small number of genuinely useful listicles and avoid building an SEO or LLM strategy around declaring our own product number one.

The original rumor is partly true. Listicles have received an unusually large share of citations for commercial AI queries. Self-promotional pages still get cited. Experiments have even shown that first-party content can introduce a lesser-known business into AI answers where it previously never appeared.

The loophole is much less attractive today than it looked at first.

Google AI Overviews can cite our page while recommending competitors. Putting ourselves first has not been shown to cause better Google performance once other factors are controlled. AI citations are unstable. Independent mentions have a much stronger relationship with AI visibility than publishing lots of pages on our own domain. And ChatGPT has already cut listicles' share of its citation mix dramatically after one model change.

So we would absolutely consider publishing something like “12 Best Email Marketing Tools for Shopify Stores” if we genuinely sell an email tool, know the category well and can produce a comparison that is more useful than what already exists.

We would test every product, make our commercial interest obvious, use criteria that can be checked, explain where competitors are stronger and give ourselves first place only when the evidence really supports it.

We would not manufacture 200 pages where every investigation somehow concludes that our company wins.

A good listicle can bring Google traffic, influence an LLM and help a human make a purchase decision at the same time. Those pages should survive changes in how ChatGPT retrieves information because their value does not depend on the loophole.

That is the version worth publishing now.