Google has made clear that spam rules now cover attempts to influence its AI-generated answers, not just old-fashioned blue-link rankings.
Google is putting the AI search industry on notice. The company has updated its Search spam policies to say that manipulation of generative AI responses in Google Search can be treated as spam, including attempts to shape what appears in AI Overviews and AI Mode.
That may sound like a small wording change. It is not. For years, search marketing has revolved around getting pages to rank higher. Now Google is saying the same enforcement logic applies when the prize is not a higher link, but a better position inside an AI answer. That matters because startups, agencies and content tools are already selling generative engine optimization as the next acquisition channel.
As Search Engine Roundtable reported on May 15, Google changed the opening language of its spam policy to cover efforts to manipulate Search systems into featuring content prominently, including attempts to manipulate generative AI responses. Google Search Central also says violations can be detected through automated systems and, when needed, human review, with consequences that include manual actions, lower rankings or removal from results.
Traditional SEO was never simple, but at least the object was clear. You wanted a page indexed, understood and ranked. AI search changes that bargain because the answer itself becomes the interface. If Google summarizes a topic, names a few sources and answers the user directly, visibility is no longer only about where a link appears. It is about whether the machine repeats, cites or trusts your material.
That has created a predictable rush. Some companies are trying to help brands become more visible in AI Overviews, AI Mode, ChatGPT, Perplexity and other answer engines. The legitimate version of this work is familiar: publish useful content, make product information clear, use structured data properly, build authority and make sure crawlers can understand the page. There is nothing inherently wrong with that.
The risky version is different. It treats AI systems as targets to be gamed. That can mean producing pages designed mainly to be quoted by an AI answer, stuffing content with unnatural instructions, creating synthetic consensus across low-value sites or using prompt-injection-style language intended to nudge a model. Google has now made the line more explicit: optimizing for users is one thing, manipulating AI output is another.
This will change the economics for SEO and content startups. A pitch built around guaranteed AI visibility is going to carry more risk if the methods look like spam. Clients may still want results, especially when paid acquisition is expensive and organic traffic is under pressure, but the cost of aggressive tactics is no longer limited to a lost ranking experiment. It can become a policy problem.
Google is also protecting its new front door
The update lands at a sensitive moment for Search. Google has spent the past two years pushing AI answers deeper into the search experience. Its May 2025 explanation of AI Overviews and AI Mode said AI Overviews use a customized Gemini model alongside Search ranking systems and the Knowledge Graph, with links meant to help users explore the web. AI Mode goes further, handling longer and more conversational queries.
That makes spam harder to define in practice. If a site buys links to rank a weak page, the old policy machinery knows what to look for. If a page includes language that appears normal to a human but is meant to steer a generative response, the evidence is murkier. Google can say the rule applies, but enforcement will depend on whether its systems can reliably distinguish clear information architecture from manipulation.
There is also a competitive tension here. Google is both the referee and a player in the new search market. Publishers and startups want visibility because AI answers can absorb attention that once went to websites. A recent arXiv study on Google AI Overviews, published May 13, found AI Overviews appeared on 13.7% of a large set of trending queries and on 64.7% of question-form queries during its measurement window. It also found that nearly 30% of cited domains did not appear in the co-displayed first-page results.
That finding cuts both ways. On one hand, AI Overviews may surface sources that traditional ranking would not have put in front of the user. On the other hand, it means businesses cannot simply map old SEO positions onto the new AI layer. The same study said 11.0% of decomposed AI Overview claims were unsupported by cited pages, which shows why bad actors have an incentive to feed the system material that looks authoritative enough to be pulled into a response.
For startups, the practical lesson is straightforward. Do not build a growth plan around tricking the answer box. Build assets that can survive scrutiny: original data, clear product pages, expert commentary, support documentation and content that answers real customer questions without hiding the commercial motive. That is less flashy than promising instant AI citation wins, but it is more durable.
The next thing to watch is enforcement. Google has now written the rule into the policy layer. The market will find out soon enough whether that language becomes a real deterrent, or whether AI search manipulation simply becomes the next cat-and-mouse business built around whatever Search cannot yet see.
Also read: Anthropic is turning enterprise AI into a workflow fight • UnitedHealth is turning AI use into a workplace metric • MiniMed is pitching diabetes care as the next self-driving system
