The AI Vendor Pitch That Almost Tanked a Major Beauty Brand

A few months ago, one of our clients pulled us into a status call with a question we've been hearing more often than we'd like: "Our parent company is bringing in a vendor to help us with AI search visibility. Can you take a look at what they're proposing?"

The proposal sounded reasonable on the surface. The vendor specialized in "agentic SEO" and "AI search optimization." They had a deck full of confident projections about traffic gains and a methodology that involved creating dedicated subdomains optimized for AI crawlers. A kind of parallel content layer designed specifically to get the brand cited in ChatGPT, Perplexity, and Google's AI Overviews.

The pitch was that human shoppers would never see these subdomains. They were "for the bots." The brand's main site would stay exactly as it was. The vendor would handle the AI-facing content separately, optimized for how LLMs read and synthesize information.

Our client's instinct was that something felt off, but they couldn't articulate why. So they sent it to us.

It took us about ninety seconds to recognize what was actually being proposed: cloaking. The textbook definition of one of the oldest and most aggressively penalized SEO violations in Google's spam policies.

If our client had moved forward, the consequences could have been catastrophic. Not just for the AI visibility goals the vendor promised. For the brand's entire organic search presence. Cloaking violations can result in complete removal from Google search results, the kind of penalty that takes years to recover from, if recovery is even possible.

The vendor either didn't understand what they were proposing, or they did understand and were betting their clients wouldn't catch it before the contract was signed. Either possibility is alarming.

This isn't a one-off story. This is the AI services market right now.

The Vacuum That AI Hype Created

Two things are happening simultaneously, and they're producing a market environment that rewards bad actors.

The first is that AI-driven discovery is genuinely changing how customers find products. ChatGPT, Perplexity, Gemini, and Google's AI Mode are all surfacing brands and products in ways that bypass traditional search behavior. Brand teams know they need to be ready for this shift. The pressure to "do something about AI" is real, urgent, and coming from the C-suite.

The second is that very few people on the brand side actually understand what AI optimization means in practical terms. The terminology is new. The best practices are still being established. Most marketing leaders are getting their information about AI visibility from the same vendors who are selling it to them, which creates an obvious and uncomfortable conflict of interest.

Into that vacuum, a wave of vendors has appeared. Some of them are doing legitimate work. Many of them are repackaging questionable tactics under new acronyms and selling them to brands who don't have the technical expertise to evaluate the proposals themselves.

The cloaking-style pitch we caught isn't unusual. It's representative. We've seen variants of the same approach pitched to multiple clients over the past six months. The branding is always sophisticated. The promised results are always specific and impressive. And the underlying tactic almost always violates either Google's quality guidelines, the AI platforms' content policies, or basic ethical standards for representing a brand to customers.

What Bad AI Visibility Tactics Actually Look Like

The vendors selling these tactics don't call them what they are. They call them "AI-first content infrastructure" or "agentic SEO frameworks" or "structured discovery layers." But underneath the language, the same handful of approaches keep surfacing.

Bot-only subdomains and cloaking. Creating content specifically for AI crawlers that human users never see. This violates Google's quality guidelines and risks complete delisting from search. Some vendors frame this as "optimizing for different audiences" or "structured data delivery." It's cloaking, regardless of how it's labeled.

Synthetic review and content generation at scale. The March 2026 Google Core Update was a significant update aiming to improve search quality by rewarding original, first-hand content and penalizing low-quality AI-generated, "filler" content. It specifically penalized sites lacking editorial oversight. Many sites, especially those utilizing excessive "AI-first" strategies, reported significant traffic drops. Focus on experience and authority: content displaying first-hand experience (high E-E-A-T) saw better rankings, while "rephrased" or thin content was hit hard.

Off-site content networks for "authority signals." Building external content sites that link back to your brand to manufacture topical authority for AI systems to recognize. This is link scheme territory, another well-documented Google quality violation that can trigger manual penalties.

Schema markup that doesn't match page content. Stuffing schema with claims, attributes, or relationships that aren't reflected in the actual content. Search engines and AI systems both validate schema against content. Mismatches don't get rewarded, they get filtered or flagged as deceptive.

Promised access to "private AI databases" or "exclusive AI training partnerships." No such databases exist in the way these vendors describe them. The major AI platforms don't have private partnership tiers that brands can buy into for guaranteed visibility. When a vendor claims they can get your products "into ChatGPT's recommendations," they're either misrepresenting the technology or planning to use tactics they aren't disclosing.

Aggressive structured data injection without architectural review. Adding schema markup at scale across an entire site without auditing the underlying content model. This produces broken or invalid markup that signals untrustworthy data to crawlers, which is worse than no markup at all.

Why These Tactics Are Particularly Dangerous Right Now

The asymmetry between potential upside and potential downside on these tactics is severe.

The upside, if any of them work, is short-lived. AI platforms are actively iterating on detection and trust signals. The cloaking tactic that surfaces a brand in ChatGPT today is the tactic that gets that brand demoted or excluded six months from now when the platform updates its content evaluation and spam policies. Even when these approaches work, the gains are temporary.

The downside is structural. Search penalties from Google spam violations don't expire when the vendor contract ends. They persist. Recovery requires months or years of remediation work, often with permanent losses in domain authority that never fully come back. For an enterprise ecommerce brand whose organic traffic represents a significant share of revenue, a manual penalty isn't a marketing setback, it's a financial event.

And the platforms being optimized for are not stable targets. The way ChatGPT surfaces products today is different from how it did six months ago, and it'll be different again six months from now. Tactics that gamed the system at one point are exactly the patterns the platforms learn to filter against. Brands that built their AI visibility on those tactics are the ones who'll find themselves invisible after the next model update.

Meanwhile, the legitimate work that produces durable AI visibility (clean structured data, comprehensive product information, accurate schema, content that genuinely answers customer questions) produces gains that compound. The same investment that helps you show up in ChatGPT today will help you show up in whatever AI platforms emerge next year. Foundational quality scales. Tactics don't.

How to Evaluate an AI Vendor Pitch

If you're being pitched on AI visibility, AI search optimization, or any variant of "AI-first" content services, here are the questions worth asking before you sign anything.

Show me exactly what you're proposing to do, in plain language. If the vendor can't explain their tactics without using "agentic," "AI-first," "next-generation," and "proprietary methodology" in every sentence, that's the whole review. Real technical work can be described in concrete terms. Vague language about "optimization layers" and "structured discovery infrastructure" is usually hiding something.

Will the content you create be visible to human users on our main site? If the answer involves "different versions for different audiences" or "subdomain-based delivery" or "schema-only enhancements," you're being pitched cloaking or thin content. The content shown to AI systems should be the same content shown to humans. Period.

Are these tactics consistent with Google's quality guidelines? Don't accept "yes" as an answer. Ask the vendor to walk you through specifically which sections of Google's guidelines apply to their approach and how their work complies. If they can't, they don't understand the guidelines, or their work doesn't comply. Either way, you have your answer.

What happens to our domain if Google updates its quality systems against tactics like these? Real vendors will acknowledge platform risk and discuss how their work mitigates it. Vendors selling questionable tactics will deflect or claim that their approach is "future-proof," which is meaningless. Nothing in SEO or AI optimization is future-proof. The right answer is "we work within published guidelines, so platform updates that target violations don't affect work that complies."

Can you show me before-and-after data from clients in similar verticals, with timelines? Watch for vendors who only show short-term traffic gains. Sustainable AI visibility work shows compounding returns over six to twelve months, not spike-and-drop patterns that suggest the gains are tactical rather than structural.

How do your tactics interact with our existing SEO performance? AI optimization that's done correctly improves traditional SEO performance because the underlying work is the same: clean data, structured content, comprehensive information. Vendors who treat AI optimization as separate from or in tension with SEO are usually doing it wrong.

What's the off-ramp if this isn't working? Reputable vendors have clear remediation paths if their work doesn't produce results. Vendors selling tactics they shouldn't be selling typically have lock-in mechanisms like content that lives on their infrastructure, schema that can't be cleanly removed, ongoing dependencies that make leaving expensive.

Where Real Strategic Value Comes From

Here's the part of this that doesn't get talked about enough: AI is genuinely making execution easier across the board. The cost of producing content, generating schema, building landing pages, and optimizing technical infrastructure is dropping fast. That's real, and it benefits brands.

But execution capacity isn't the bottleneck most enterprise brands are facing. The bottleneck is judgment.

The brands that get the most value from AI in 2026 won't be the ones who deployed the most tactics. They'll be the ones who avoided the tactics that hurt them. Who said no to the vendors selling shortcuts. Who invested in foundational quality instead of platform-specific gimmicks. Who had a trusted technical advisor in the room when proposals like the one our client received hit their inbox.

These brands ultimately ask the question: "Is the work we're doing adding real value to my audience, or am I just doing this so that AI will see it?"

That's where strategic value lives now. Not in execution, which is increasingly commoditized. In the experience, analysis, and foresight required to evaluate what's being sold to you, see around corners, and protect your brand from decisions that look smart on paper and disastrous in practice.

For our client in this story, the strategic value of our involvement wasn't measured in what we built. It was measured in what we helped them say no to. The cloaking proposal got shut down. The vendor relationship was renegotiated. The brand's organic search presence, and the millions of dollars in revenue it generates, was protected.

That kind of moment isn't rare anymore. We've had similar conversations with multiple clients over the past two quarters. The vendors change. The tactics evolve. But the pattern is the same: ambitious AI promises, sophisticated branding, and underneath, an approach that would have caused serious harm if it hadn't been caught.

The Real Test of an AI Partner

If you're evaluating AI vendors, agencies, or technical partners right now, the most useful test isn't what they promise to build for you. It's whether they'll tell you when something is a bad idea.

Vendors who only know how to sell you services have an incentive to say yes to every project. Partners who understand the technical and reputational stakes have an incentive to push back when a request is risky, premature, or just wrong for your brand.

The agency that talks you out of a tactic that would have hurt you is worth more than the agency that builds whatever you ask for. The advisor who flags a vendor's cloaking pitch is worth more than the one who would have helped you implement it.

This isn't an argument against trying new things or moving fast on AI. It's an argument for moving fast with the right people in the room. The brands pulling ahead in AI search aren't doing more, they're doing less of the wrong things, faster.

If your team is being pitched AI visibility services and you want a technical second opinion before you sign, that's the kind of conversation TAG is happy to have. We've reviewed enough of these proposals at this point to recognize the patterns, and we'd rather help you avoid a bad decision than build something on top of one.