Quick Answer
Google's spam policies define spam to include attempting to manipulate generative AI responses in Google Search, and that wording was added on 15 May 2026. The June 2026 spam update then refreshed detection worldwide, applying globally and to all languages. Ordinary answer engine work is untouched, because what the policies reach is any tactic whose result depends on a signal you made rather than earned.
Most of the AI optimization pitches reaching founder inboxes were written before Google put any of this in writing. They still promise guaranteed mentions inside ChatGPT, a text file you drop at your root, a formatting trick that gets your page quoted back to a buyer.
Google's own documentation now answers nearly all of it, and the answers are duller and far more useful than the pitches. A few of those tactics are genuinely spam. Many more are simply ignored, which is a different problem and a cheaper one to fix.
This covers the exact policy sentence and the date it actually changed, the four tactics Google's own guide tells you to skip, and the one report that shows whether any of your work moved.
Google's Spam Policies Now Cover AI Answers
Google's spam policies now reach AI answers directly. The policy page opens by defining spam as techniques used to deceive users or manipulate Search systems into featuring content prominently. It then gives two examples of doing that: ranking content highly, and "attempting to manipulate generative AI responses in Google Search."
Spam includes "attempting to manipulate generative AI responses in Google Search." — Google's spam policies for Google web search
Where that sentence sits matters more than the sentence itself. It is in the definition of spam, not in a footnote about one tactic, which means there is no separate AI ruleset bolted on the side.
The categories that governed blue links now govern what appears inside AI Overviews and AI Mode. If a tactic was already a violation, pointing it at an AI answer does not make it a new discipline. It makes it the same violation with a new target.
1. The Definition Changed in May 2026, Not July
Google's documentation changelog dates the clarification to 15 May 2026. The entry is headed "Clarifying that spam policies apply to generative AI responses in Google Search." Google's stated reason was to make clear the policies cover all of Google Search.
The same day, Google published its guide to optimizing for generative AI features. Anything you have read placing this change in July is repeating a secondhand timeline rather than the changelog.
The date is worth getting right because it tells you what the change was. Google clarified an existing rule; it did not write a new one, and the distinction decides whether your last twelve months of work needs re-examining.
2. Scaled Content Abuse Is the Policy That Bites
Of the named categories, scaled content abuse is the one most AEO programs run into. Google defines it as many pages generated for the primary purpose of manipulating search rankings and not helping users, whatever produced them.
The AI guide applies it to a tactic being sold right now. Google describes query fan-out as the related queries a model generates to gather more information. Building separate content for every possible variation, it says, violates the scaled content abuse policy when the point is to manipulate rankings or AI responses.
That is the concrete version. One near-identical page per fan-out phrasing, published so a model trips over your brand more often, is the shape the policy describes.
You can test your own program in one question: if Google published exactly how the retrieval worked tomorrow, would this work still be worth doing? The common mistake is answering that question about the content and not about the reason it exists.
3. Borrowed Authority Falls Under the Site Reputation Policy
The other one founders get pitched is renting a strong domain. Google's site reputation policy covers third-party content published on a host site mainly because of ranking signals that host earned from its own first-party content.
Note what it does not say. Having third-party content is fine; the policy turns on whether the content is there mainly to borrow the host's standing. A guest post you wrote because that audience is yours is not the same object as a paid slot on a news domain chosen for its authority.
Google also lists freelancers and white-label services as examples of separate entities, so "our team wrote it" is not the defense people assume it is. The test is why the content sits on that domain rather than yours.
The June 2026 Spam Update Refreshed Detection Worldwide
The June 2026 spam update was enforcement, not new rules. Google's Search Status Dashboard records the release at 09:03 Pacific on 24 June 2026 and the rollout as complete on 26 June 2026, applying globally and to all languages.
Read the entry itself and what is absent is the story. It names no policy, because a spam update improves how Google detects the policies that already exist.
When Google genuinely adds a policy, it says so separately. In April 2026 it added back button hijacking to the malicious practices policy and published a Search Central blog post explaining it, which is what a new rule looks like.
So the two events run on different tracks. Google's spam policies picked up explicit AI wording in May, and the detection systems behind them were refreshed in June. Neither is a quality judgment about your topic.
Legitimate AEO Compared With Answer Manipulation
The line between legitimate answer engine work and manipulation is not about technique. It is about where the signal comes from, and every tactic that survives scrutiny produces something a reader would value even if no model ever read it.
| Question to ask | Legitimate AEO | Answer manipulation |
|---|---|---|
| Where the signal comes from | Work you did and published | Signals created to be counted |
| If Google explained the mechanism | Nothing about the plan changes | The tactic stops working |
| Who benefits besides you | The reader who lands on it | Nobody |
| Policy it touches | None | Scaled content abuse, site reputation policy |
| What a vendor can promise | Process, publishing, reporting | A guaranteed mention |
The last row is the one that ends most sales calls early. A vendor can commit to what they will do and what they will show you, but nobody can commit to what a model says next Tuesday.
Run your current plan down the first column. If a line item only makes sense under the second, Google's spam policies already describe it and you are paying for exposure rather than results.
Four AI Tactics Google Says You Can Ignore
Most of what gets sold as AI optimization is not against Google's spam policies at all. It simply does nothing, which is why the vendor can keep selling it without anything visibly going wrong.
1. LLMs.txt and Other Machine-Readable Files
Google's AI optimization guide is explicit on this one. You do not need new machine-readable files, AI text files, markup or Markdown to appear in Google Search, because Search does not use them. The 15 June 2026 changelog entry adds that the files will not affect your visibility or rankings either way.
Keeping an llms.txt file for another service is fine. Paying for one as an AI visibility strategy is buying a file that Google ignores.
2. Chunking Your Pages Into Smaller Pieces
There is no requirement to break content into small pieces so a model can understand it. Google says its systems handle multiple topics on a page and surface the relevant part.
Shorter or longer pages can both work depending on the subject. Google's own framing is that there is no ideal page length, which removes the main justification for restructuring a library that already reads well.
3. Rewriting Pages in a Voice for Machines
You do not need to write in a particular way for generative AI search. Google says its systems understand synonyms and general meaning, so you do not have to capture every phrasing someone might use.
This is the expensive one, because it usually arrives as a rewrite of pages that were working. The rewrite costs you the specific voice that made the page worth citing in the first place.
4. Chasing Mentions You Arranged Yourself
Seeking inauthentic mentions across the web is listed in the same guide as something that is not as helpful as it looks. Be precise here, because the internet has this one wrong. There is no "inauthentic mentions" entry in Google's spam policies, and we checked the live page rather than a roundup of it.
It sits with the ineffective tactics, not the punishable ones. The reason it fails is structural. Google's generative features lean on the same ranking systems that favor good content, and on the same systems that block spam.
What Actually Earns a Place in AI Answers
Google's answer to what works is almost aggressively unexciting. Its guide says creating content people find unique, compelling and useful will likely influence your presence in generative AI search, in the long run, more than anything else in the document.
The guide spells out what that means in practice:
- A unique point of view, from someone who has actually done the thing, rather than a summary of what is already online.
- Non-commodity content. Google's own contrast is "7 Tips for First-Time Homebuyers" against a first-hand account of waiving an inspection.
- Organization a human can follow, with real headings and sections.
- A technical structure that lets the page be crawled, indexed and shown with a snippet.
None of that is a trick, and none of it is new. It is the same work that made a page rank before answers appeared above the links. Nothing in Google's spam policies touches any of it.
Better Marketing treats content marketing for tech companies as exactly this: the founder's own knowledge, written down. Our AI SEO service starts from that page, then the crawling, then the measurement. It turns out to be the shape Google describes.
Search Console Now Measures Generative AI Impressions
You can now see your own AI visibility without buying anyone's estimate of it. Google launched Search Generative AI performance reports in Search Console on 3 June 2026, and its announcement carries a note saying the insights reached all websites worldwide on 31 August 2026.
The reports show impressions, the pages that appeared, countries, devices and dates for generative AI features in Search and Discover. That is a first-party number for AI Overviews and AI Mode, which is more than any dashboard could honestly offer a year ago.
There is a sharper reason to prefer it. Google's spam policies include machine-generated traffic, which covers sending automated queries to Google, explicitly including scraping results for rank-checking without permission.
So a tool that builds its numbers by scraping Google is not only fragile, it is doing something the policy names. We wrote about what to ask an AI visibility tool before paying for it, and the first question is where the data comes from. An AI SEO service that cannot answer that about its own reporting is selling you a slide, not a measurement.
Where a Narrow Scope Beats a Bought Reputation
This is where the advice gets uncomfortable, because the method has a real limit. If you launched last month with no domain authority and no published work, everything above is slow, and Google's spam policies close off the fast version.
The honest response is not to buy manipulation. It is to shrink the target until your specific knowledge is enough to win it.
Pick one category, one buyer problem, one set of proof points, and publish the best explanation of that narrow thing that exists anywhere. Getting cited with no audience works on exactly that principle, and it is how content marketing for tech companies pays off for a company nobody has heard of yet.
Say the quiet part too. If your runway needs customers this quarter, none of this arrives fast enough, and a B2B demand generation agency running founder-led outreach is the more honest answer than any AEO retainer. Search work is the thing you start now so that next year is cheaper.
Earn the Citation Instead of Forcing It
The manipulation era of AI search ended before it properly started. Google's spam policies name AI answer manipulation directly, the June 2026 update refreshed the systems that enforce them, and most of the remaining tactics were never doing anything.
Publish what only you can write, keep it crawlable, and read the Search Console report instead of a vendor's deck. If you want help building that into a system rather than a list of tasks, Better Marketing runs the search side and the B2B demand generation agency side together, so trust and pipeline build at once.