Hi everyone!

For the past two weeks we've been moving in the same direction. First Lily Ray said "being visible isn't enough, you need to be recommended." Last week, with Amazon's agent ads, we reached "recommended isn't enough either, you need to be buyable." This week, another link in that chain arrived, and I think it's the sneakiest one: if an AI agent can't get into your site and act on the user's behalf, none of the above matters.

Here's the deal: an SEO asked John Mueller a question. "Gemini can use a computer now, it can browse the web on a user's behalf. Will your quality principles, like 'images should provide a satisfying experience,' evolve in this agent era?" Mueller's answer is two-layered, and both layers matter. First: most principles will stay the same, because a site that's useful for humans is generally useful for agentic browsers too. Second, and this is the real news: "some new basics will come into play, such as not blindly blocking agentic browsers."

So Google is effectively saying: locking AI agents out of your site is about to become an SEO mistake. It sounds very technical, but it's actually a strategic warning. That's what we'll cover this week.

Grab your coffee, let's dig in.

On June 26, 2026, in response to an SEO's question, Google's John Mueller made a critical comment about agentic browsers and search quality principles. The question was: now that AI tools like Gemini can autonomously navigate the web and retrieve information on a user's behalf, will Google's quality principles change? Mueller's answer is clear: "I expect most principles will remain the same.

A website that's useful for users will generally also be useful for agentic browsers." But the critical addition: "Some details will undoubtedly evolve (and new basics, such as not blindly blocking agentic browsers, will come into play), but in the end, it's still users." So Google draws a line between content quality and technical accessibility: even if your site meets every quality standard, if you won't let AI agents through the door, you're shooting yourself in the foot.

The SEO who asked the question hit a very real point: in Gemini 3.5 Flash, "computer use" is now a built-in feature. So the agent isn't just reading text, it's genuinely browsing the site like a user, clicking buttons, filling forms, completing tasks. In that case the subject of a "satisfying experience" changes: sometimes the one having that experience isn't a human, it's an information agent.

Mueller's answer kept the core logic intact: a good site is a good site, whether a human or an agent looks at it. But with the "don't blindly block" warning, he underlined a new technical reality. Because many sites block AI bots wholesale, to reduce server load or to protect their content. Mueller is saying that block can soon backfire on you.

Search Engine Journal explains it with an old example: years ago, when the nofollow tag came out, some sites blocked off their own About Us pages to funnel PageRank to "important" pages, and ended up hurting themselves. The agentic-browser situation may follow the same pattern: a technical decision made for one reason ends up with unintended SEO consequences.

This completes the missing link in the chain we've been discussing for two weeks. Being visible, being recommended, being buyable, we kept assuming content and authority were what's at stake. But Mueller steps back and asks: can the AI agent physically get into your site? If your robots.txt or your firewall rejects agentic crawlers, neither your content quality nor your authority does anything. If the agent can't get through the door, it can't read you, understand you, recommend you to the user, and it certainly can't act on your behalf.

There's a fine balance here. On one side is the publisher logic we discussed last week with the beehiiv and Cloudflare example: "I'll block the bot that takes my content and gives me no traffic in return." On the other is Mueller's warning: a blind block can make you invisible in AI's eyes. So the blocking decision is now a strategic one that needs fine tuning; the wholesale "shut it all off" approach is becoming risky.

I took this comment seriously, because it puts a finger on a rarely-discussed but increasingly critical area: agent accessibility. For years, "technical SEO" meant Core Web Vitals, mobile usability, crawl budget. Now a new layer of technical SEO is emerging: whether AI agents can reach your site and complete tasks on it. Call it "agent readiness."

Last week, discussing Amazon's agent ads, I said "AI now buys on your behalf." Mueller's comment gives the technical foundation for that picture: for that purchase to happen, the agent has to get into your site, see the product, read the price, add it to the cart. A brand whose site blocks the agent isn't at the table in the agentic-commerce era.

My real takeaway: the definition of SEO is expanding. We used to say "let Google's bot into your site"; now we say "let the AI agent into your site too, and make its job easy." An agent-friendly site is tomorrow's ranking site.

This week, open your robots.txt and your server/CDN settings and ask: which AI crawlers am I blocking? Am I wholesale rejecting GPTBot, Google-Extended, ClaudeBot, PerplexityBot, and especially agentic crawlers? If so, reconsider that decision. Separate "the bot I don't want to be visible to" from "the agent I need to be visible to." Instead of blocking blindly, set a deliberate access policy. And make sure your site is browsable and actionable by an agent: clear buttons, open form fields, machine-readable product data.

What happened? Google began rolling out the June 2026 spam update at around noon ET on June 24, and the update finished rolling out at 2pm ET on June 26, taking two full days. This is the second spam update of 2026, and by Barry Schwartz's read it felt a bit bigger than the March 2026 update. The critical point is tied to a May policy change: Google had explicitly stated that its spam policies now cover efforts to manipulate generative AI responses in Search. So tactics built to game AI Overviews or AI Mode, buying citations or artificially inflating them, now fall under the same rules as older spam tactics.

What it means for SEO: This draws a formal line under the "buy the citation, hack your AI visibility" approach spreading across the GEO world. Resorting to artificial means to appear in AI answers now carries the risk of a manual action or automated demotion. Google has put traditional spam and AI spam in the same frame.

My take: I read this update alongside last week's Lily Ray study and this week's other news. They all say the same thing: AI visibility is earned with deserved authority, not with manipulation. Don't panic, but be cautious. If you dropped in the spam update, as freelance SEO consultant Shushrita M. put it on Search Engine Journal: a sudden decline doesn't automatically mean your content is "bad"; the right response is to identify which page types, queries, and directories were affected, then look for a consistent pattern. Recovery starts with diagnosis, not panic. I always say the same: in core and spam updates, analysis first, then action.

Practical step: Look at the June 24-26 window in Search Console. If pages dropped, find what they have in common. Especially if you used artificial methods for AI visibility (bought mentions, inflated citations, scaled-up comparison pages), you're in the risk pool. Rework those pages to add real value.

What happened? Search Engine Land published a notable analysis this week examining a Google LLM patent. The 2023 patent describes how AI systems could "understand" brands, companies, products, and people from websites and public data. Google calls this process building a "deep, holistic characterization" of an entity. So the system isn't just indexing content; it interprets the information, extracts relationships, generates summaries, and tries to understand the entity behind the content.

What it means for SEO: This shows SEO's focus shifting from "optimizing the page" to "explaining the entity, that is, your brand, to AI." We used to optimize individual documents, titles, keywords. Now the real issue is whether Google and other AI systems can build a clear answer to "who are you, what do you do, who do you serve?" The patent defines an entity broadly: people, companies, places, objects, concepts.

My take: This sits at the heart of the entity SEO approach I've described for years in our geo-master work and in this newsletter. Wikidata, consistent NAP data, the Knowledge Graph, schema, author authority: they all came down to one job, teaching AI who you are. Now Google's own patent confirms it. It also ties directly to this week's headline on agent access: even if the agent can get into your site, if it can't clearly understand who you are, it still can't recommend you. Access is a precondition; being understandable is the real work.

Practical step: Build your brand's "entity file." When you search your brand on Google, does a Knowledge Panel appear? Do you have a Wikidata entry? Is Organization and Person schema set up correctly on your site? Are your name, address, and description consistent across platforms? If the answer is "no," start there so AI can understand you.

What happened? SimilarWeb's research this week reveals an important blind spot in how we measure AI visibility. Based on a U.S. desktop panel across the finance, travel, and beauty sectors, 55.9% of the downstream traffic a site gets after a user is exposed to a ChatGPT recommendation comes via branded search. So when AI mentions your brand, people often don't click a link; they leave and search your name directly on Google. SimilarWeb also found that AI-recommended brands saw 2.5x more site visits.

What it means for measurement: This is a critical warning for GEO measurement. As Aleyda Solís put it: "AI influence can happen without a click, and this is why measuring AI Search impact only through 'AI referral traffic' is not enough." Current attribution models have a blind spot: AI-influenced demand often arrives as Search and Direct traffic, not as "AI referral." So "5 visitors came from ChatGPT" in your analytics may be showing a tiny fraction of the real impact.

My take: This study confirms, with concrete data, the "we're measuring AI visibility wrong" thesis I've been making for months. Most brands try to measure GEO success by "clicks from AI," see a low number, and panic. But the real impact is hidden in branded search. When AI recommends you, the user remembers your brand and searches you directly. So I now tell my clients: track your branded search volume as an indicator of your AI visibility. If your branded search is rising, you're in good shape in AI, even if you can't fully see where the traffic comes from.

Practical step: Start tracking the volume of your branded queries (queries containing your brand name) in Search Console over the last three months. Read it alongside your AI visibility. If your branded search is climbing, AI is talking about you. Add this metric to your monthly reports; don't get stuck on the "AI referral" number.

What happened? The AI website builder Manus offers a built-in SEO feature: the system automatically generates a prerendered, static HTML version of your site for search engine bots; it creates canonical URLs, robots.txt, and dynamic sitemaps on its own; it gives an SEO health score; and with an "Optimize with Manus" button it auto-fixes titles, descriptions, and keywords using AI.

What it means for measurement: This is part of a broader trend: a chunk of technical SEO is being embedded into AI tools and automated. That's good, because prerendering, canonicals, and sitemaps were always a barrier for non-coders. But there's a caveat: automated SEO provides the "technical foundation," not the "strategy." Manus can generate a meta description for you, but it can't generate who you are, which intents you serve, which entity you are.

My take: I read this news as both an opportunity and a trap. The opportunity: the boring, repetitive part of technical SEO is being automated, which frees us to focus on the real work, strategy and content. The trap: the "the tool does my SEO for me" illusion. No, the tool builds the technical foundation; but this week's headline and patent show exactly that the real issue is teaching AI who you are and letting it into your site correctly. No button can do that for you. Use AI tools to speed up the technical foundation, but keep the strategy to yourself.

Practical step: If you use an AI website builder like Manus or a similar automated SEO tool, don't leave the auto-generated titles and descriptions as is. Take what the tool produces as a draft, then add your own entity identity, your own voice, and your own intent targeting on top. Automation lays the foundation; you create the difference.

What happened? Last week, during Prime Day, Amazon introduced a new ad format called Alexa+ Agentic Ads: the user sees the ad, asks Alexa a question, compares options, and completes the purchase without going anywhere, without leaving the conversation. It launched on Echo Show devices with partners like Papa Johns (food ordering) and Beck, Jill Scott, and Omar Courtz (concert tickets). It's the first major concrete ad example of "agentic commerce," and it explains why this week's Mueller comment matters so much.

What it means for AI: These two pieces of news complete each other. Amazon shows that AI can now buy on a user's behalf. Mueller shows that for that purchase to happen, the agent has to get into the brand's site and act. So in the agentic-commerce era, two things are true at once: your product data has to be machine-readable, and your site has to be browsable and actionable by an agent.

My take: My caution about agentic commerce continues, but it gets more concrete week by week. A purchase this frictionless raises a trust question: if the agent orders on my behalf, how do I know that recommendation is really in my interest and not in the interest of the highest-paying advertiser? That's why I always come back to the same place: in the agentic-commerce era, the winning brand won't be the one that advertises most, but the one AI and the user trust most. Because a user won't open their wallet on the recommendation of an agent they don't trust. Trust is the currency of this new world.

Practical step: Check how clearly machines can read your product and service data: structured data (schema), current price, stock status, delivery info. Then combine it with the headline's step: beyond that data, make sure an agent can actually get into your site and complete the transaction. Clear data, open door, strong trust. Those are the three legs of agentic commerce.

There's one lesson this week: in the agent era, the AI agent now browses the web on your behalf and even buys for you; but if you blindly block that agent from your site, neither your content quality nor your authority nor your entity identity does anything. John Mueller said it with a technical rule, the Google spam update drew a line under manipulation, and SimilarWeb showed the real impact is hidden in branded search. They all land in the same place: open the door, explain clearly who you are, earn the trust.

The core of this week is short: in the AI-agent era, opening the door matters as much as guarding it. The agent now browses the web on your behalf, and even buys for you. But if you blindly block it, neither your content quality nor your authority nor your entity identity works. Open the door, explain clearly who you are, earn the trust.

Here's the good news, same as always: this is an era where the most trusted and most understandable brand wins, not the 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: