Google has drawn a harder line around AI search than most vendors wanted. In July 2026, its spam policies now explicitly define spam to include attempting to manipulate generative AI responses in Google Search. That matters because the June 2026 spam update just finished rolling out globally, and a wave of “AI optimization” pitches is still telling founders that shortcuts are strategy.
The short version is simple. Legitimate AEO is still allowed. Manipulation is not. If a tactic depends on fake signals, manufactured mentions, or pressure on systems instead of earning relevance, it belongs on the wrong side of the line.
The new line
Google’s spam policies page now includes a sentence that changes the frame for founders. Google says spam includes “attempting to manipulate generative AI responses in Google Search.” You can read the policy on Google’s own site here, in the section on spam policies, at developers.google.com/search/docs/essentials/spam-policies.
This is not a vague cultural signal. It is a policy statement. Google is no longer treating AI answer manipulation as an adjacent concern, it is naming it as spam.
That should immediately narrow the acceptable set of AEO tactics. If your strategy is built on making your brand appear more often in AI summaries by gaming the system, the policy now speaks directly to that behavior.
The June update
The timing matters. Google also completed the rollout of the June 2026 spam update between June 24 and June 26. Search Engine Journal reported that this was the second spam update of 2026, and that no new policies were announced with the rollout, only enforcement of existing spam systems. See Search Engine Journal’s report.
That distinction is important. Spam updates are not core updates. They are designed to improve how Google detects and handles spam, not to reward one topic area over another. If your process looks like manipulation, the update is relevant whether you call the work SEO, AEO, GEO, or something else.
In practice, this means Google has two things in motion at once, a policy that names AI answer manipulation as spam, and a spam system that was just refreshed globally. That is not a friendly environment for shortcuts.
What is spam
For founders, the useful question is not whether a tactic sounds clever. It is whether it depends on creating false signals. The main categories that now look clearly unsafe are easy to identify.
- Coordinated fake mentions or reviews, created to seed AI answers with fabricated authority.
- Parasite or site-reputation plays, where low-quality content is placed on strong domains only to hijack trust.
- Scaled AI content farms, published to flood the index and influence AI Overviews through volume rather than value.
Those tactics share the same flaw. They are attempts to influence systems by manufacturing evidence instead of earning it.
That is also why many “GEO hacks” collapse under policy scrutiny. If the pitch depends on deception, the tactic is not an optimization method, it is a liability.
Snake oil
There is also a separate pile of advice that does not rise to the level of spam, but still does nothing.
- LLMs.txt files, Google has said they will not help or hurt. They are not a ranking lever, and they are not a compliance shield. See the July 2026 roundup at Search Engine Roundtable.
- Guaranteed ChatGPT mentions, no one can promise that. If a vendor does, they are selling certainty they do not control.
- Markdown-for-LLMs pages, Google still treats HTML as the standard for indexing and rendering. Reformatting content for machines does not replace useful content for humans.
Founders should be precise here. Some vendor claims are merely useless. Others are risky. A good budget should avoid both.
What works
Legitimate AEO still exists, and it looks boring for a reason. It is built on material that large models can trust because users and publishers already trust it.
- Earned citations from credible third-party sources.
- Original data, especially if it is specific to a market or workflow.
- Clear direct-answer content, written so the answer is easy to extract.
- Consistent entity signals across your site, profiles, and earned media.
None of that is manipulation. It is evidence.
This is where Better’s view remains unchanged. If your content is genuinely useful, it can earn citations in search systems and in AI systems. If it is not genuinely useful, no layer of formatting will save it.
Measure now
The measurement story is improving, which matters because too many AEO vendors sell confidence without instrumentation. According to Search Engine Roundtable, Search Console AI performance reports are rolling out. That gives founders a better way to see how AI surfaces are interacting with their content, instead of relying on a service provider’s claim deck. See the July 2026 Google Webmaster report.
If you are early stage, this is enough to get disciplined.
- Identify the pages that answer your buyer’s highest-intent questions.
- Check whether those pages are clear enough to cite without interpretation.
- Review whether your brand is appearing in AI-related performance reports.
- Remove any tactic that only works if the system is fooled.
The last step is the most important. Measurement should expose reality, not justify theater.
Zero authority
If you are pre-launch or still have weak domain authority, the temptation to cut corners is higher. That is also when bad vendors get persuasive. They know you want visibility fast, and they know you cannot yet compete on reputation alone.
The right answer is not to buy manipulation. It is to narrow scope.
Focus on one category, one buyer problem, and one set of proof points. Publish the best explanation in your market. Seed it with real examples, not synthetic ones. Then earn references from places your buyers already trust.
When you do not have authority yet, the temptation is to simulate it. Google has now made that the wrong bet, technically and policy-wise.
FAQ: Is AEO spam
Will AEO get my site penalized?
No, not by itself. AEO is not the problem. Tactics that attempt to manipulate generative AI responses are the problem. Google’s policy language is about manipulation, not about writing helpful content.
Is optimizing for AI Overviews allowed?
Yes, if you are doing it through legitimate publishing, clear answers, and earned citations. It becomes risky when the goal shifts from helping users to gaming AI outputs.
Do I need LLMs.txt?
No. Google has said LLMs.txt will not help or hurt. Treat it as optional documentation at best, not as a growth strategy.
Can vendors guarantee mentions?
No credible vendor can guarantee durable mentions in AI answers. Anyone promising that should be treated as selling control they do not have.
Takeaway
The manipulation era of GEO died before it started. Google has now named AI answer manipulation as spam, refreshed its spam systems, and said LLMs.txt does nothing. That leaves a narrow but workable path.
Earn the citation. Publish useful work. Measure what happens. Ignore anything that depends on fooling the system.
If you want the safer route, build the kind of content Google can trust and buyers can verify. That is the only AEO strategy that looks durable now, and the only one that belongs in a revenue system.