A Citation Is Not a Recommendation

You wrote a "Top 10 Best [Your Category] Software" post, put yourself at number one, and called it content marketing. For fifteen years that worked. As of this spring, Google's AI Overview reads your list, cites your page as a source, then tells your buyer to go sign up with the five competitors you mentioned in it.

This article proves one thing: when you rank yourself in your own "best" list, you are writing the exact script an AI reads aloud to your buyer, and that script names everyone but you. The fix is not a better list. The fix is to stop ranking yourself and start being the brand other people's lists name. A self-promotional list reads to the model like an ad, so it gets cited as a source and never trusted as a recommendation.

Google is citing your "best" page and recommending the brands you listed under yourself

Danny Goodwin reported the pattern in Search Engine Land on June 18, 2026, under a headline that says the quiet part out loud: Google AI Overviews cite self-serving listicles but recommend competitors 69% of the time. The number comes from a study by Lily Ray at Amsive, who analyzed 100 B2B "best [category] software" queries across three dates this year, on April 15, May 15, and June 8, pulling data through Ahrefs Brand Radar.

What she found: Of the 100 queries, 80 triggered an AI Overview. In those 80, self-promotional listicles got cited 323 times. And in 224 of those 323 citations, Google quoted the brand's own page as a source but recommended somebody else. That is the 69%. Two out of every three times your self-ranked "best" page shows up in an AI answer, it shows up working for your competition.

Translate that into a sales meeting. A buyer asks Google which software to pick. Google pulls a sentence from your page to build the answer, then lists four other vendors as the recommendation. Your buyer never sees your name as the suggestion. They see it as a footnote. You paid a writer to produce the citation that helped four competitors close a deal you were trying to win.

The cleanest example in the study involves a learning platform. For the query "best LMS for selling courses," Google cited Oasis LMS's own page, then recommended Kajabi, Thinkific, LearnWorlds, and Teachable. Every one of those four names appeared inside the Oasis article. Oasis wrote the list. Google used the list. Google handed the recommendation to the four companies Oasis named, and not to Oasis.

This was not a one-off in a single category. Ray found the same citation-but-not-recommendation pattern across help desk software, task management, survey tools, CRM, and SEO software. Different industries, same outcome. The page gets read; the brand gets skipped.

You were trained to do this, and the timing of your traffic drop is not a coincidence

The self-ranked "best" page was not a mistake when it was invented. It was the correct move for a decade. You wanted the keyword "best CRM software," so you wrote a post targeting it, and because it was your post, you put yourself first. Google rewarded the page with rankings, the rankings sent traffic, and the traffic converted. The logic held all the way up until it didn't.

The break has a date. According to Ray's study, organic declines for sites leaning on self-ranked "best" pages began around January 20, 2026, then accelerated during Google's May 2026 core update. If your traffic from those pages started slipping in late January and fell off a ledge in May, you were not imagined. You were early to a pattern that is now documented.

The reason the old tactic broke is that the reader changed. For ten years the reader of your "best" page was a human who could see your logo at the top, understand you were the publisher, and mentally discount the bias. The reader now is a model, and the model handles your bias differently. It does not discount your self-ranking and then trust you anyway. It treats the self-ranking as a signal that the page is promotional, lifts the factual claims it can use, and looks elsewhere for who to actually recommend. I've written before about the gap between being mentioned by AI and being believed by it, and this study is the cleanest proof of that gap I have seen. Your "best" page buys the mention. It cannot buy the belief.

There is a second reason businesses keep doing this, and it is more uncomfortable. The self-ranked list is easy. You control it. You can publish it this afternoon without asking anyone's permission, without earning anything, without a single outside party agreeing that you belong on the list. Earning a spot on somebody else's list is slow, partly outside your control, and impossible to fake. So people keep choosing the easy thing that no longer works over the hard thing that does.

The data says the recommendation now lives on pages you don't own

If Google is not recommending the brands that rank themselves, where is the recommendation coming from? The study answers that too. For "best" queries, the most-cited domains included Forbes, Reddit, and YouTube, with Reddit citations rising sharply across the three measurement dates.

Read that against the 69% figure and the strategy writes itself. Google trusts the page where the recommendation came from someone who was not selling. A Reddit thread where six strangers argue about CRMs carries more recommendation weight than your polished page where you happen to win. Search Engine Land framed the stakes plainly in its editorial.

"A citation is not a recommendation. Your content can appear in an AI answer while helping competitors capture the visibility that matters most."

Search Engine Land, June 18, 2026

Sit with the cost of that for a second. Earlier reporting on this shift found that some B2B and SaaS brands lost 30 to 50% of visibility after leaning on self-ranked "best" pages. For a company that gets half its pipeline from search, losing a third to half of its AI visibility is not a content problem. It is a revenue problem that shows up two quarters later when the pipeline that fed sales quietly stops refilling.

There is also a legal edge most owners have not clocked. The FTC Consumer Review Rule is triggered by an undisclosed material connection, which is a plain way of saying you have a stake in the outcome that the reader cannot see. You control the content and you wrote the ranking, but you present that ranking as if an independent judge handed it down. The reader has no way to tell the scoreboard is owned by one of the contestants. That hidden ownership is the problem, not the ranking itself. The fix is not complicated. You either add a clear, visible disclosure that says in plain words that you publish this list and your product appears on it, so the reader can weigh the bias for themselves, or you remove your own self-placement entirely so the page is no longer crowning its own publisher. Either move turns a page that can read as a deceptive endorsement into one that is honest about who is talking. The same page that has stopped working for AI can start working against you in a different room entirely.

What to do Monday instead of writing another list you rank yourself in

Start by auditing what you already have. Pull every page on your site with "best," "top," or a ranking in the title where your own product appears in the ranking. For each one, run the query it targets through Google and look at the AI Overview. If Google is citing your page but recommending competitors, you have found a page actively working against you. Either rewrite it so it stops ranking you against the field, or remove your self-placement entirely and let it function as honest comparison content. This is a one-afternoon job for whoever manages your site.

Second, redirect the energy you spent ranking yourself toward earning third-party mentions. The recommendation now lives on Forbes, on Reddit, on YouTube, on review platforms, on independent roundups you do not control. You cannot bribe your way onto those, and you should not try. What you can do is become genuinely worth naming, then make sure the people who write those lists know you exist. The case that brand mentions, not backlinks, now drive AI search visibility has been building for a year, and the Amsive data is the receipt for it.

Third, stop trying to outrank the communities. A lot of owners see Reddit climbing in citations and decide they need to beat it with a better page. You will not beat it, and the trying wastes the budget. The smarter move is to earn an honest presence inside those communities and platforms. The full case for why you should stop trying to outrank Reddit and earn presence on the platforms AI already trusts sits in its own piece.

Fourth, if you have an agency, ask them one question and watch how they answer. The question is: "For our top ten money queries, is Google's AI Overview recommending us or just citing us?" An agency that knows what it is doing will already have the answer or will get it to you within a week with screenshots. An agency that responds by talking about keyword rankings and domain authority is measuring the old game and billing you for it. Citing and recommending are different outcomes, and any partner who cannot tell you which one you are getting is not watching the thing that determines your pipeline.

Fifth, fix your own comparison content so it earns trust instead of broadcasting bias. If you publish comparison pages, publish ones you would still publish if your product were not on the page. Name where competitors genuinely win. Be specific about who you are not for. A page that openly says "if you need X, go with this other tool" reads to a model as less promotional and more trustworthy than one where you sweep every category. The goal is to stop sounding like the ad you have been writing.

Measure recommendation, not citation, and ignore the metric that will lie to you

The trap waiting for you is the citation count. It is easy to pull, it goes up when you publish more, and it tells you almost nothing about whether anyone is being told to buy from you. A brand can watch its citations climb while its recommendations crater, which is precisely the 69% problem in Ray's data. If your dashboard celebrates "AI mentions" and stops there, your dashboard is lying to you with a straight face.

Measure the split instead. For your top money queries, track two separate numbers: how often AI cites you, and how often AI actually recommends you. The gap between those two is the entire story. Pull this monthly, not once, because a single check tells you nothing reliable. I've written about why asking ChatGPT one time whether it recommends you is worthless and how to measure recommendation as a rate across many runs. Treat it like a hit rate, not a yes or no.

On timing, give changes a full quarter before you judge them. The damage in this study unfolded across months, from a January slide to a May acceleration, and recovery runs on the same clock. If you strip self-rankings off your pages in July and check the AI Overview in August, you will not see much. Check the recommendation rate across your money queries in October against your July baseline, and you will have a real read. Anything faster is noise you will misread as signal.

Watch one more number that has nothing to do with AI dashboards: the share of your inbound that mentions finding you through a third party. You do not need software to track this. Add a single required question to your intake form or your sales-call notes, the oldest one in the book, "How did you hear about us?" Then once a month, tally the answers that name a third party, a forum, a roundup, a video, a peer who pointed them your way, and divide that by your total inbound for the month. Write the number down. The first month is just a baseline and it will not tell you anything on its own. What you are watching for is the trend. After you strip the self-rankings off your pages, watch whether that third-party share climbs over the following quarter. When prospects start saying they saw you recommended on a forum, in a roundup, in a video, that is the recommendation working in a place you do not control, which is the only place it counts now.

This distinction is one piece of a larger system, and I cover the rest in my AI search visibility guide.

Frequently Asked Questions

Should I delete my "best" pages entirely?

Not automatically. Delete or rewrite the ones where you rank your own product against the field, because those are the pages Google cites while recommending your competitors. A genuine comparison page that names where rivals win and where you do not fit can stay, because it reads as honest rather than promotional. The test is simple: would you publish this page unchanged if your own product were not on it? If not, it is an ad wearing a list costume, and it is the kind of page that began losing visibility in January 2026.

If Google cites my page, doesn't that still help my brand?

A citation gets your name into the answer, but the Amsive study found that in 69% of citations of self-promotional lists, Google recommended a competitor instead of the cited brand. A citation is the footnote; the recommendation is the sale. Being named as a source while four rivals get suggested as the choice is not a win, it is your content doing free work for the other side. Track how often you are recommended, not just cited, or you will mistake exposure for demand.

How do I get onto the third-party lists AI actually trusts?

You earn it, which is slower than writing your own list and the reason most companies avoid it. Forbes, Reddit, and YouTube were among the most-cited domains for "best" queries in the study, with Reddit climbing fast, and none of them can be bought cleanly. Become genuinely worth naming in your category, build honest presence in the communities your buyers already trust, and make sure the people who compile independent roundups know your product exists. The work is real, but it produces the one thing your own list no longer can: a recommendation a model will repeat.

For fifteen years the smartest people in the room told you to write the list and put yourself on top, and they were right, until January, when the audience for that list stopped being a person and became a machine that reads your self-praise as a confession of bias. The companies winning AI search right now are not the ones with the best self-ranked pages. They are the ones other people keep naming. Go look at your own "best of" post, the one you are proud of, and ask whether you wrote a recommendation or an advertisement. The model already knows the answer, and it is telling your buyers every day.