Warning: AI Search “Vacuums” Enable Fraudsters to Hijack Brands Without GEO and EEAT
By Charles Ryder
I’ve been watching a quiet shift turn into a very loud problem: people aren’t finding brands the way we planned for. They’re asking AI engines for answers—and treating the summary as the source of truth.
That change creates what Jonathan Armstrong aptly called AI search “vacuums”: gaps where an AI system doesn’t confidently know your brand, your official support channels, your policies, or your credibility. When those gaps appear, fraudsters don’t just exploit them—they define you inside them.
This isn’t hypothetical. It’s already showing up as fake customer support numbers, impersonation pages, and scammy “help centers” getting surfaced through AI-driven results. And as AI gets cheaper, content-based fraud scales faster. If you’re treating this like “just another SEO trend,” you’re already behind.
The core risk: if AI doesn’t recognize you, someone else will volunteer an identity for you
In classic search, brand defense meant owning domain variants, managing paid search, cleaning up misleading pages, and outranking sketchy results. Those tactics still matter, but AI search changes how trust gets assigned.
Generative engines (Google AI Overviews, Perplexity, ChatGPT-style search experiences, Claude, and others) don’t simply list websites—they compose answers. That composition depends on what the system can retrieve, cite, or infer. If it can’t find strong official signals, it fills the blanks with whatever looks credible.
Fraudsters understand that dynamic better than most marketing teams.
They seed content on scraped sites, publish “support” pages with fake phone numbers, spin up plausible FAQs, or cloak pages so crawlers see something authoritative while humans get redirected. When an AI summary repeats that poisoned information, it’s not just a ranking problem—it’s AI-mediated brand impersonation at scale.
Why this problem is getting worse, fast
Two forces are compounding the risk.
1) Users are choosing AI summaries over links
A YouGov survey found that more than half of Americans prefer AI summaries to traditional search listings. That shift means fewer clicks, less cross-checking, and a stronger tendency to treat the AI’s wording as the “official” answer.
2) The economics of AI-powered fraud have collapsed
AI strategist Nina Schick highlighted a brutal reality: inference costs that were about $60 per million tokens a few years ago have fallen to roughly six cents. That’s the difference between “a scammer can try this sometimes” and “a scammer can do this industrially.”
Cheap generation lets fraudsters:
- create thousands of plausible support pages,
- spin endless variations of “official-looking” FAQs,
- localize scams by region, language, and product line,
- refresh pages faster than brands can respond.
The visibility crisis: major brands are barely present in AI answers
One of the most alarming data points I’ve seen comes from Geometriqs’ GenAI Global80 Report: average brand visibility in AI summaries was around 4%, and 1 in 5 major brands were completely absent. Financial services were even worse—about 2.9% visibility—which is basically an open invitation for investment scams, fake onboarding, and “account recovery” fraud.
Here’s what that looks like in real life:
- Your customer asks an AI engine: “What’s the support number for [Brand]?”
- The AI doesn’t have strong official signals to pull from.
- It retrieves a number posted on a scraped “help” page.
- The customer calls.
- The scammer “verifies” the account and drains it.
This pattern has been reported in the context of AI Overviews surfacing scam phone numbers for household-name brands. The mechanism isn’t mysterious. It’s an information supply chain—and brands are losing control of the upstream inputs.
The hidden self-inflicted wound: robots.txt and “AI-blocking” that creates a vacuum
There’s an uncomfortable twist: some brands are making the vacuum bigger.
Crawlers like GPTBot and Google-Extended often respect robots.txt rules. If you blanket-block AI crawlers without a plan, you can accidentally guarantee that:
- the model can’t reliably access your official support pages,
- it can’t reach your policies or documentation,
- it can’t quote your authoritative sources,
- it falls back to whatever third-party copies exist.
That doesn’t “protect your content.” It can erase your official footprint right where users now get answers.
I’m not arguing for opening everything up without thinking. I’m saying treat this like risk management, not a philosophical stance.
GEO + EEAT: the new baseline for brand safety in AI discovery
GEO (Generative Engine Optimization) isn’t a marketing nice-to-have anymore. It’s brand defense. It works best when paired with EEAT—Experience, Expertise, Authoritativeness, Trustworthiness—because generative systems (and the retrieval layers around them) are constantly deciding what’s credible.
What GEO actually means (without the buzzword gloss)
GEO is the discipline of shaping how AI engines:
- discover your brand,
- interpret your authority,
- reproduce your facts in answers.
It’s not just “rank me.” It’s “represent me correctly.”
What EEAT does in this environment
EEAT reduces the chance that AI treats your brand as interchangeable with a scraped clone. Signals that matter include:
- clearly attributed expert authorship,
- consistent “official” contact data across trusted sources,
- documented policies and support workflows,
- first-party evidence (original research, official statements, verified listings),
- and increasingly, rich media (video, demos, interviews) that’s harder to counterfeit at scale.
Practical moves I recommend to close the vacuum
If I were advising a brand team trying to reduce AI search exposure right now, I’d start with five actions:
-
Lock down “official support truth”
- Publish a canonical support page with phone numbers, domains, and escalation rules.
- Make it easily retrievable and clearly labeled as official.
-
Audit AI answers like you audit security logs
- Run recurring checks on common prompts: “support number,” “refund policy,” “login help,” “pricing,” “where to buy,” “official website.”
- Track changes weekly or daily—AI outputs drift.
-
Review robots.txt and crawler access deliberately
- Avoid a situation where scrapers are visible but your official sources aren’t.
-
Strengthen structured data and authoritative citations
- Make it easy for machines to extract verified facts.
- Ensure consistency across trusted profiles and references.
-
Treat impersonation as a cross-functional incident
- Marketing, legal, compliance, and security should share one playbook.
- The EU AI Act’s focus on AI literacy and risk management isn’t abstract—this is the kind of scenario it’s pointing toward.
The uncomfortable truth: “SEO is dead” isn’t the point—brand trust is
I’ve seen the debates: SEO isn’t dead, GEO is hype, GEO tools are wrappers. Fine. Argue about terminology if you want. The outcome is what matters:
When AI becomes the interface, your brand becomes a dataset problem. If your official narrative isn’t present, consistent, and trusted, someone else will manufacture a narrative that is.
And because AI summaries often create “zero-click” journeys, you may not even see the traffic dip that would normally warn you something’s wrong. The scam can play out entirely inside the AI answer layer.
Conclusion: record reality—or AI will invent it for you
AI search “vacuums” are the next major fraud surface, built for brand hijacking: low-cost content generation, high-trust AI presentation, and users trained to accept the summary. GEO and EEAT aren’t just growth tactics now—they’re how you prevent scammers from becoming the default voice of your brand.
If you want a practical way to start measuring and improving how generative engines represent you, use a solution built for this new visibility layer—AIuthority makes it much easier to see where you’re exposed, where you’re missing, and what to fix before someone else fills the vacuum.