Noise vs. Signal – Issue 02

Google made “tricking the AI” a spam violation. The vendors selling those tricks should worry, not you.

Late in June, Google ran a spam update, a tune-up of the automated systems that catch sites breaking its rules. It started on June 24, finished about two days later, and hit every language and region at once. Cue the usual wave of worry, with plenty of site owners who have never spammed anything in their lives spending the week refreshing their analytics.

Here is the calmer fact underneath that. A spam update is narrow. It runs on SpamBrain, Google’s spam-catching system, and it goes looking for specific violations like cloaking, scraped pages, and piles of low-value content. Barry Schwartz checked with Google and confirmed this one does not even touch link spam or the site-reputation rules. If you are not using those tricks, it is almost certainly not about you.

The part actually worth your attention happened five weeks earlier. Back in May, Google quietly added one line to its rulebook: trying to manipulate the AI answers in Search (the AI Overviews at the top of results, and the chat-style AI Mode) now counts as spam, the same as the old “black hat” tricks for gaming blue-link rankings. It was the first time Google named AI-answer manipulation as a violation in writing, and it called out specific tactics: “recommendation poisoning” and rigged comparison lists built to get a brand quoted.

That quietly reframes a whole category of sales pitch. The tactics a certain kind of vendor sells under the banner of GEO or AEO as urgent must-dos, things like stuffing “recommendation” pages, buying citations, or paying someone who claims a special line into the AI, are not a clever edge anymore. They are a way to get demoted. And on that last one Google has been blunt: no outside tool or vendor has any special access to how its AI picks answers. Anyone selling you that access is selling you a story.

One caution before you read too much into your own numbers. The May core update only finished on June 2, and this spam update arrived three weeks later, so two separate systems moved in the same month. Recovery from a spam demotion is measured in months, not days. The calm move is to wait for the dust to settle and check whether the right pages are still there, rather than making sudden changes mid-wobble.

Sources: Google’s spam-policy update naming AI-answer manipulation as spam (May 15) and the June spam update logged on Google’s Search Status Dashboard (the sources that count) · coverage from Search Engine Land and Search Engine Journal · Barry Schwartz at Search Engine Roundtable, who confirmed with Google that this update does not target link spam or site-reputation abuse. Reinforced by Cyrus Shepard’s review of 54 studies, which found classic search rank is the single strongest predictor of being cited in an AI answer.

You can finally see whether you show up in AI answers. Google put it in the dashboard you already use.

For two years, the answer to “are we showing up in the AI Overviews?” was a shrug. You could see your total impressions in Google’s reporting tool, Google Search Console, but no way to tell which slice of them came from AI. As of June 3, there is a real answer.

Google added a new report to Search Console. It shows how often your pages turn up inside three AI surfaces: the AI Overviews and AI Mode in Search, and the AI features in Discover. You can break it down by page, country, device, and date, right down to the hour.

Three caveats keep it from being the full picture. It shows impressions only, meaning how often you appeared, not how many people clicked or what they typed to get there. It is rolling out to a handful of sites first, starting in the UK, so you may not see it in your account yet. And John Mueller has clarified that a link tucked inside a “show more” is only counted once someone expands it, so a quiet number can understate you.

Bundled into the same release is a switch that lets you remove your site from AI answers entirely. It takes effect from June 17, and Google has promised it will not be held against you in normal search rankings. For the overwhelming majority of businesses, walking away from AI answers means trading real visibility for nothing, so this is a switch almost nobody should touch.

It helps to be clear on what this is not. The numbers themselves are not new. Google says AI impressions were always folded into your overall totals. What changed is that you can finally see them on their own, which is the whole difference between guessing and measuring.

Sources: Google’s Search Central announcement of the generative-AI performance reports (June 3) · John Mueller’s clarification of how impressions are counted · hands-on walkthroughs from Brodie Clark and Search Engine Journal.

The one new idea this month: your loyal readers can now pin you into their own AI answers.

Here is something genuinely new this month, and rare, because there is no trick to it. Google has a feature called Preferred Sources that lets a person star the websites they trust so those sites turn up more often. It emerged in 2025 as a way to pin your favourite news outlets to the top of the news box. This month Google brought it into the AI answers.

Now when someone who has “starred” your site runs a relevant search, you’re clearly flagged and favoured inside their AI Overview and AI Mode results, not just in the news listings. Google says more than 345,000 sites have already been starred, nearly four times the number from December.

The part that matters for you is what it cannot do. There is nothing to optimize here. You cannot buy your way in or write your way in, and no agency can game it for you. John Mueller has been clear that it sits alongside the normal ranking systems rather than jumping the queue past them. The only lever is the reader’s own choice, which quietly makes “building an audience” and “doing SEO” the same job.

Acting on it is almost embarrassingly simple. Google hands you a ready-made deeplink that opens its Preferred Sources page with your site already filled in, so a loyal reader can add you in a single click. Put it in your newsletter, add a small button next to your social links, and ask once.

One number deserves a raised eyebrow. Google says people click a Preferred Source about twice as often as other links, but it has not shown how it measured that, or whether it accounts for the obvious fact that people click sites they already like more anyway. The feature is real and worth using. That particular stat is worth quoting carefully.

Sources: Google’s announcement via product manager Duncan Osborn (rolling into AI Overviews and AI Mode from June 1) · coverage from Search Engine Journal and Semrush · SEO consultant Glenn Gabe on what it means for visibility.

Sources: zero-click and AI Overview coverage rates from Semrush and Similarweb tracking · Cyrus Shepard’s review of 54 AI-citation studies · Google’s own Preferred Sources figures · with a reminder from growth advisor Kevin Indig to start from the business question, not the latest number.