5 AEO vendor claims worth verifying — and the questions that test them

Questions this guide answers

• How do I evaluate an AEO vendor?
• Which AEO vendor claims should I verify?
• What questions to ask an AEO platform before signing?
• Is AEO a real category or vendor hype?

This article is about claims, not vendors

I want to be clear about what this essay is — and what it isn't — before the first claim.

It's about pitch patterns I keep seeing in AEO vendor conversations. It is not a list of specific platforms. I'm not naming names because I don't have to: every claim below is a pattern broad enough that you'll hear it from multiple vendors in your evaluation, sometimes including from SolCrys. The point of the essay is to give you a single question that tests each pattern, so you can quickly tell a serious operator from a marketing-first pitch.

At the end, I apply all five tests to SolCrys publicly. If we can't pass our own bar, that's worth knowing too.

Why the test set just got sharper

In May 2026 Google published an official guide on optimizing for generative AI features. Several of the tactics that AEO vendors have been pitching — `llms.txt` files, AI-specific schema, forced content chunking for Google AI Overviews, per-fan-out page production — Google named explicitly as not required and, in some cases, as anti-patterns under the scaled-content-abuse spam policy.

That doesn't mean every AEO platform was selling those tactics, but it does mean the conversation has changed. A vendor pitch that still leans on the named tactics now needs to defend itself against published Google guidance. A vendor pitch that doesn't go near them is — at least on that dimension — credible by default. The five tests below sit on top of that floor.

Claim 1: "We optimize for AI Overviews and AI Mode"

Almost every AEO pitch you'll hear includes this line. It is also the easiest claim to test, because Google's May 2026 guide is unusually specific about what "optimizing for AI Overviews" actually means: do the SEO fundamentals, write non-commodity content with a distinctive point of view, keep technical accessibility clean. There is no AI-specific lever.

What to verify: ask the vendor what specifically changed in their methodology after Google's May 2026 guide landed. If they have a substantive answer — "we retired our llms.txt deployment recommendation," or "we reframed our schema work as visible-content alignment rather than AI citation lift" — they're paying attention. If they give a generic answer ("we already aligned" with no specifics), or if their published content still recommends Google-AI-specific tactics that Google has now named as not needed, the claim is marketing rather than operational.

Most legitimate AEO platforms can describe what their position on AIO and AI Mode is and how they update it when the source-of-truth (Google's own guidance) moves. That update behavior is the diagnostic, not the claim itself.

Claim 2: "We monitor 50+ AI engines"

Engine coverage is a real differentiator, but the headline number on its own is hollow. There are perhaps five or six AI surfaces where most B2B and consumer brands' buyers actually spend real attention — ChatGPT, Perplexity, Google AI Overviews and AI Mode, Claude, Gemini, plus retail RAG assistants like Amazon Rufus, Walmart Sparky, and ChatGPT Shopping for ecommerce. Beyond that, monitoring becomes long-tail and the signal-to-cost ratio gets worse fast.

What to verify: ask the vendor to segment their engine list by your buyers' real demand for your specific category. Which five engines drive the majority of measurable buyer attention for you? How does the vendor know? If the answer is "we cover them all, so it doesn't matter," they're optimizing for a headline number, not for your buyers' actual answer surface. If the answer is "these five for your category, here's the data," they're operating.

Coverage breadth is useful as a hedge against future engine emergence. It is not useful as a substitute for depth on the engines that matter today.

Claim 3: "We auto-publish content to win citations"

Automatic content publishing is increasingly part of AEO platform marketing. Done well, it's a workflow feature: the platform identifies a gap, drafts content, routes it through a brand approval step, and ships only after a human has signed off. Done poorly, it's a brand-risk pattern dressed up as productivity.

What to verify: ask the vendor to walk through the governance step. Who reviews the draft before it ships? What's the platform's failure-mode case — a real example where the auto-published draft was wrong, what went wrong, and how the system caught it. If the vendor describes a one-click "approve and publish" UI without a substantive review path, the auto-publish claim is a brand-risk vector for you, regardless of how good the drafting model is.

The right framing here is that auto-drafting is fine; auto-publishing without governance is the failure mode. The platform's posture on that distinction tells you whether they're optimizing for demo velocity or for production usability.

Claim 4: "Our AI agent fixes the gap for you"

Agent pitches are everywhere in 2026 marketing technology, and they're often vague about what the agent actually does. The marketing version is "the agent identifies the gap, runs the playbook, and the gap is closed." The honest version names the agent's failure modes and who reviews when it fails.

What to verify: ask the vendor what the agent's failure modes look like and who reviews them. Specific failure modes: how does the agent behave when an engine API rate-limits it; what happens when the brand fact the agent needs is missing from the source of truth; what's the rollback if the agent ships something the brand didn't want. A vendor that can describe specific failure modes is running an actual production agent. A vendor that says "the agent just works" is selling a demo.

Related: our own marketing agent infrastructure positioning essay covers the broader pattern of vendors using "agent" as a positioning label rather than a product capability. Worth a read in parallel.

Claim 5: "Guaranteed citation lift"

AEO measurement is non-deterministic by nature. Engines have their own retrieval logic, their own sampling behavior, and their own update cycles. Citation share for any given brand can move based on factors entirely outside any AEO platform's control — engine model updates, competitor publications, third-party media coverage. Against that backdrop, "guaranteed" lift is either a marketing claim untethered from how the measurement actually works, or a conditional claim with terms the vendor hasn't disclosed yet.

What to verify: ask the vendor for the worst-performing customer case in their last twelve months and what went wrong. If they can describe a specific case where the lift didn't materialize, name the diagnostic that surfaced it, and explain what they changed in response, they have a real practice. If they can't or won't, the guarantee is marketing.

We have a separate essay on the engineering reasons AEO measurement is more like survey design than search analytics — when published, it will go deeper on why "guaranteed lift" is structurally hard to defend. For now: ask for the failure case.

The SolCrys self-test

Here are our honest answers to the same five tests.

What to do with this

Run the five tests against your current AEO vendor — or the ones you're evaluating — in your next conversation. The point isn't to disqualify anyone; it's to surface which pitches are operational and which are marketing. A pitch that passes the tests is worth your time. A pitch that doesn't deserves either a follow-up question or a redirected budget.

If you'd like the longer form of the SolCrys self-test methodology, the evaluate-aeo-platform-methodology-checklist is the operational version of these five tests, in checklist form for procurement teams. Apply it to us with the same standard as anyone else.

FAQ

Related guides