Hi everyone!

For the past few weeks I've kept repeating one sentence in this newsletter: being visible in AI isn't enough. This week, the hardest, most concrete proof of that sentence arrived. Lily Ray sat down and measured it, and showed that in some cases being visible isn't just insufficient, it actually hurts you. Your own content can hand the win to your competitor instead of you.

Here's the deal: for years, one of the most popular AI tactics has been writing a "top 10 best brands in the industry" listicle and putting your own brand at number one. Lily Ray examined 100 B2B queries across three different dates and found something surprising: Google cites these listicles as sources in its answers, but drops the brand that crowned itself number one out of the recommendation. On top of that, in 69% of cases, the competitors that brand named in its own listicle get recommended. So the "we're the best" content you wrote yourself turns into a vote cast for your rival.

This lays bare the difference between citation and recommendation. The two are assumed to be the same, but they aren't. A citation means "you were used as a source." A recommendation means "you're the one suggested to the user." And the one that drives the business outcome is the second.

On top of that, Google published new research this week on how it detects AI junk content (slop), OpenAI brought improved health responses to free ChatGPT, and Bing opened a brand-new AI reporting panel. There's a lot to cover.

Grab your coffee, let's dig in.

Bottom line up front (BLUF): In her study published on June 17, 2026, Lily Ray examined 100 B2B "best [category] software" queries across Google's AI Overviews at three separate dates (April 15, May 15, June 8). The finding is striking: when a brand publishes its own "best" listicle and ranks itself number one, Google cites that article as a source in its answer but leaves the brand out of the recommendation 69% of the time. Worse still: the competitors you named in that listicle get recommended. So your own content becomes a vote for your rival. The critical distinction here: being cited (citation) and being recommended (recommendation) are two different things, and the one that drives the business outcome is the recommendation.

Lily Ray used Ahrefs Brand Radar data to pull each query's AI answer and the sources it cited, then split every answer into two separate metrics. First, citation: does the brand's own "best" article appear among the answer's sources? Second, recommendation: is the brand actually one of the options recommended inside the answer?

The result is clear. Across the 80 queries that surfaced an AI Overview, when a brand's own self-promotional listicle got cited as a source, that brand was left out of the recommendation 69% of the time. Across the 100 queries tracked over three months, 74 returned an answer that cited a self-promoter's listicle but left that brand out of the recommendation. The ones recommended were always the same type: the established, authoritative category leaders that plenty of other independent pages already recommend.

Lily Ray gives an example: on the query "best LMS for selling courses," Oasis LMS is cited repeatedly, both in the body and in the sidebar, but isn't recommended. Even though it crowns itself "number one" in its own article. Meanwhile the competitors it named in that article, Kajabi, Thinkific, LearnWorlds, Teachable, all make it into the recommendation.

This study takes the "being visible isn't enough" theme we've been discussing in recent weeks to its sharpest point. Google appears to have decoupled what it cites from who it recommends. The citation decision is fed by content; the recommendation decision is fed by how much the rest of the web talks about you, links to you, that is, by real authority signals.

Lily Ray shows this with numbers. The recommended brands have far more referring domains and far more AI Overview and ChatGPT mentions than their cited-but-not-recommended counterparts. So calling yourself "the best" changes nothing; others calling you "the best" is what changes it.

There's also a new move from Google: it now adds a disclaimer to some "best" queries. When Lily Ray asked for "the best SEO experts," AI Overviews warned that the field is "saturated with self-proclaimed experts." So Google is aware of self-promotion and is warning users against it.

This study both delighted me and confirmed, with a bitter clarity, something I've been saying for years. I've always said AI visibility isn't bought from a tool or a trick, it's earned. The shortcut of "I'll rank myself number one in my own listicle and be done with it" worked in the short term because it exploited a gap. Now that gap is closing.

My field observation backs this up too. The most common mistake I've seen in content audits these past months is brands stuffing their own pages with self-praising, unsubstantiated superiority claims. This kind of content erodes credibility even for a human reader, remember Burson's "facts convince, adjectives don't" finding from last week. Now Lily Ray shows that the same holds for AI, and even harder: that unsubstantiated "we're the best" content wins the game for your competitor, not you.

My real takeaway is this: we need to stop treating the citation count as a success metric. There are plenty of people celebrating "ChatGPT cited me." Lily Ray is saying a citation on its own means nothing, and can even mislead you. What matters is being recommended. And recommendation comes not from praising yourself, but from the web praising you.

Do two things this week. First, if you have self-promotional listicles where you crown yourself "number one," review them. Unless you're an established, authoritative brand, those pages are probably hurting you by carrying your competitors into the recommendation. Second, shift your energy from praising yourself to getting others to praise you: independent reviews, real customer cases, third-party mentions. That's where AI's recommendation decision is shaped.

What happened? Google researchers published a paper on June 19 describing a new way to detect coordinated AI-generated spam ("slop"). The system is called the Scalable Cluster Termination System (S-CTS). Although the research focuses on video spam, it also covers text-based AI-generation detection. The key point: instead of evaluating content one piece at a time, the system looks for the "structure" of an attack. That is, the mass reuse of the same template, the same narrative pattern. Google also explains that for text it uses Sentence-BERT (S-BERT), a technique for detecting the mathematical "fingerprint" that AI-generated text leaves behind.

What it means for SEO: This is an admission that "content-level quality filtering is no longer enough." Spam has scaled so much that Google is moving from evaluating individual pages to detecting the production pattern and the infrastructure behind them. For you, the meaning is: producing "functionally identical" content in thousands of small variations to slip past the filter is increasingly risky. Google is no longer looking at the single page, but at the cluster that produced it.

My take: I took this research seriously, because it draws a technical line under the era of flooding search results with scaled AI content. I've said for years: use AI not to multiply content, but to deepen it. The existence of techniques like S-BERT shows that the logic of "fill the template, multiply, publish" can be caught mathematically. An original voice, real experience, your own sentence structure: these are now a defense not just for the reader, but for the filter too.

Practical step: Read your content asking "could a thousand other pages have written this same sentence with the same template?" If the answer is yes, that content is at risk. Add your own data, your own example, your own sentence. Step away from the template.

What happened? Google's May 2026 core update completed on June 2; as of this week the SERPs have settled. Lily Ray's study offers a data point here too: sites that had been using self-promotional listicles at scale since January started losing organic visibility, and that decline accelerated with the May core update. What's more, the loss often wasn't confined to a single folder but spread across the entire domain.

What it means for SEO: Two updates pointing the same way in a row isn't a coincidence. Google can penalize tactics that skirt the edges of its policies not just on that page, but across the whole site. Self-promotional listicles, scaled AI content, comparison/alternative page factories: these all sit in the same risk pool.

My take: This shows why the "push the limit, and if you get caught only that page drops" math is wrong. Spam signal accumulating in one folder drags down the trust of the entire domain. My advice after a core update is the same: move with analysis, not panic. If you dropped, look at which pages dropped and what they have in common. In this update, the common thread was often "self-praising, unsubstantiated, scaled-up" content.

Practical step: Open the last three weeks of data in Search Console and list your 10 biggest-dropping pages. If they're "we're the best" self-promotion or scaled-up pages, that's likely the source of the problem. Either rework that content to add real value, or take it down.

What happened? On June 16, Microsoft refreshed Bing Webmaster Tools' AI reporting with four new features: Intents, Topics, Citation Share, and Compare. It's rolling out in preview globally. This panel lets you see for which intents and topics, and at what share, your site is being cited across Copilot and Bing's AI surfaces.

What it means for measurement: AI visibility measurement is slowly entering official tools. On the Google side, Search Console's AI reports are coming; now Microsoft offers a concrete metric like "citation share." That's good, but with a caveat: as we discussed in this week's headline, citation share on its own is a misleading metric. Your citation share in Bing might be high, but the real win is being the one recommended.

My take: I'd advise not neglecting the Bing side, because part of ChatGPT's web searches are fed by Bing's infrastructure. So being visible in Bing can indirectly affect your ChatGPT visibility too. The new panel is free; at the very least, take a look and see which topics you're being cited for. But don't fall in love with the "citation share" number; read it as a starting signal, not as the final win.

Practical step: If you don't have a Bing Webmaster Tools account, open one this week and add your site. In the new AI reporting panel, look at which topics and intents you're being cited for. If there are important topics where you come up empty, that means you have a content gap there.

What happened? Peec AI analyzed 37,804 AI responses across five AI engines (ChatGPT, Gemini, Perplexity, Google AI Mode, AI Overviews) and found this: how exactly users word their prompt matters less than you think. About 88-92% of the ways people phrase the same commercial need turn out to be very close in meaning. Brand visibility stays stable as long as the intent stays the same, even when the words change. But there's one exception: the middle of the funnel, that is, unbranded commercial discovery queries, is the zone most sensitive to wording changes.

What it means for measurement: This is a good antidote to the "I have to track every possible prompt variation" panic. No, you don't. If the intent is fixed, largely the same brands surface even when the words change. But format has an effect: prompts like "give me a list" or "compare" surface up to 20% more brands than open-ended questions do.

My take: This study rescues GEO from the "chase every keyword" frenzy and refocuses on what matters: intent. My takeaway is this: instead of chasing prompt variations, understand the real intents of your audience and build your content around those intents. Tracking more variations makes sense specifically for the commercial queries in the middle of the funnel, but a few representative queries are enough for the top and bottom. When you measure AI visibility, read each engine separately; because, as Peec showed, the engines don't behave the same.

Practical step: When tracking your AI visibility, split the queries you monitor by funnel stage: roughly 25% top-of-funnel, 50% middle, 25% bottom. Spend most of your budget on the middle-of-funnel commercial queries, which are the most sensitive to wording changes.

What happened? OpenAI announced that GPT-5.5 Instant, the default model for free ChatGPT, now rivals its paid frontier models on health questions. The company says the rate of health responses flagged for at least one possible factuality issue fell 71% over two months. In one comparison, a panel of physicians even rated GPT-5.5 Instant's responses higher than doctor-written ones. But an important caveat: all of these results rest on OpenAI's own evaluations, with no independent or peer-reviewed validation.

What it means for AI: Health is one of the categories with the highest visibility rate in AI answers, and more than 230 million people a week ask ChatGPT health questions. That huge, free audience now gets the answer directly in ChatGPT, without clicking through to a source. For publishers producing health content, that means even more "zero-click" pressure.

My take: The lesson here isn't just for the health sector. OpenAI says "our model beats doctors" based on its own testing, with no independent validation. That sits right at the heart of the believability issue we discussed last week. An AI saying a claim doesn't mean the claim is true. Whatever sector you're in, if users now get the answer from AI, it's your job to verify that answer's accuracy and, where needed, correct it with your own authority. Trusting blindly what AI says is healthy for neither the user nor the brand.

Practical step: Ask ChatGPT 3-5 critical questions about your sector and check the accuracy of the answers with your own expertise. If you see anything wrong or missing, address that topic correctly, with evidence and up to date, on your website. AI only carries it into its answer if it finds a clear source on your site.

What happened? A notable analysis ran on Search Engine Land this week: the line between paid and organic visibility is fading, because the same AI (Gemini) drives both. PMax, AI Max, AI Overviews, AI Mode: there's one engine behind all of them. So the AI optimizing your ad and the AI evaluating your organic content are the same; it's reading the same user, in the same moment, with the same intent.

What it means for AI: This is the end of the "the paid team and the organic team live in separate worlds" mindset. According to the author, "teaching" your brand to Gemini through organic improves your paid performance through the same mechanism. One investment, two outputs. Because when Gemini serves an ad in real time, it grounds it against the organic context; that is, the same knowledge graph, the same index, the same model.

My take: This thesis is a bet, not hard proof, and the author says so plainly. But the logic is sound and it matches my field observation. Whether AI "understands and trusts" your brand is becoming decisive on both the organic and paid sides. If Gemini doesn't recognize and trust your brand, neither organic content nor ad settings will save you. So SEO, GEO, and even advertising increasingly come down to one job: making your brand trustworthy and understandable in the eyes of AI.

Practical step: If you have two siloed teams reporting paid and organic separately, sit them at the same table this week. Pull from your ad data which cohort-intent-profit combinations actually convert, then build organic content around those combinations. Think of the two as a single loop.

There's one lesson this week: being cited is one thing, being recommended is another. Calling yourself "the best" doesn't work in AI, it even hands the win to your competitor; because the recommendation decision is made not by your praise, but by the web praising you. Lily Ray showed it with numbers, Google reinforced it with slop detection, and they all land in the same place: earned, evidence-backed authority wins.

The core of this week is short: stop praising yourself, get others to praise you. Lily Ray's study, Google's slop detection, even the merging of paid and organic, all point the same way. AI rewards earned authority; not inflated, self-appointed authority.

Here's the good news, same as always: this is an era where the brand that's genuinely trusted wins, not the one shouting loudest. If you'd like to keep up with these moves in real time, SEOs Diners Club is where I share the weekly read; if you want to go deeper with strategy on specific markets or verticals, that's what Stradiji does for enterprise clients in the US, UK, and Australia.

If this newsletter has been useful to you, you can add Stradiji.com as a Preferred Source in Google with this link. One click, five seconds, and we'll meet more often inside Google's AI answers.

See you next week with more.

Mert Erkal

I genuinely enjoy sharing what's moving in SEO and AI search with you every week. If you find the work useful, you can keep it going by buying me a coffee. ☕ Click 'Buy Me a Coffee' to contribute to the knowledge-sharing. Let's build more together.

Mert Erkal is a digital marketing and SEO expert with 15+ years of experience. As the founder of Stradiji, he provides consulting on SEO strategy, GEO (Generative Engine Optimization), conversion rate optimization, and AI integration to global brands across the US, UK, Australia, and beyond.

More information: