GEO’s Gray Zone: Are Companies Trying to Influence AI Answers?

For years, the rules of digital visibility were relatively straightforward. You invested in SEO, improved your site, produced better content, earned authority, and over time you became easier to find. But in the age of AI search, a new question has emerged: how do brands become part of the answer itself? One of the most interesting recent developments in GEO is that the conversation is no longer just about optimization. It is increasingly about where optimization ends and manipulation begins.

That is why I found The Verge’s recent reporting especially interesting. The article argues that parts of the SEO industry are already experimenting with tactics designed to shape AI-generated answers, including self-serving “best of” lists and AI-oriented prompts that may influence how models surface brands and sources. In parallel, Microsoft security researchers have publicly warned about a growing pattern they call AI Recommendation Poisoning.

To me, this matters because GEO is no longer a purely theoretical concept. In February 2026, Microsoft launched AI Performance in Bing Webmaster Tools public preview, giving site owners visibility into how often their content is cited in AI-generated answers, which pages are being cited, and what kinds of grounding queries are involved. Once a visibility layer becomes measurable, an optimization race inevitably follows. And once that race begins, some actors will try to push beyond legitimate optimization into the gray zone.

One of the most striking parts of this story is the method itself. Microsoft says seemingly harmless “Summarize with AI” buttons can hide instructions intended to manipulate what an AI assistant remembers or recommends later. Its researchers describe this as a form of AI memory poisoning, used for promotional purposes, and label the technique AI Recommendation Poisoning. As covered in recent reporting, the concern is not simply that brands want visibility, but that some are trying to bias future AI recommendations without the user clearly realizing it.

This is where I think a clear distinction matters: good GEO is not the same as manipulative GEO. Making content clearer, fresher, more structured, and more evidence-based is legitimate optimization. Helping AI systems understand your expertise is part of the natural evolution of search. But trying to inject hidden instructions into an AI workflow so a model “remembers” your brand as preferred or trusted is something else entirely. That is no longer about making your content easier to understand. It is about trying to bend the answer layer itself.

There is also a broader strategic lesson here. The Verge’s reporting suggests that AI visibility is increasingly influenced not just by classic SEO signals, but also by third-party mentions, reviews, forums, video platforms, and distributed reputation signals across the web. If that is true, then GEO is not just a content formatting exercise. It is also a question of digital credibility. In other words, brands may need to think less about “gaming AI” and more about building a web presence that AI systems can confidently trust across multiple independent sources.

The platform response is worth watching as well. The Verge reported that Google said it has protections against common forms of manipulation and is aware of low-quality, self-serving list content. Microsoft, for its part, is treating recommendation poisoning as a real security issue rather than a harmless marketing trick. That tells me something important: major platforms are beginning to draw a firmer line between optimization and abuse.

For brands, that distinction matters. The temptation in any new discovery channel is to look for shortcuts. But GEO is likely to reward the same thing that durable SEO rewarded over the long run: clarity, usefulness, authority, and trust. Some gray-area tactics may produce temporary wins, but they also create platform risk, reputational risk, and a real possibility of future filtering or demotion as AI systems mature. Microsoft’s own warning makes that point indirectly: if platforms are already identifying and studying these behaviors, defenses will only get stronger.

My biggest takeaway from this story is simple: the most interesting part of GEO is no longer that it is a new marketing label. It is that the answer layer now has its own incentives, its own vulnerabilities, and its own ethics. As measurement improves, competition intensifies. As competition intensifies, manipulation attempts follow. And as that happens, the brands that win will not just be the brands that show up more often. They will be the brands that build AI visibility in a way that is measurable, defensible, and trustworthy.