Can our SEO agency fix our AI visibility?

No, and it's worth understanding why before you renew the retainer. A good SEO agency can make your site easier for a model to crawl, easier to parse, and easier to match to a query. None of that decides whether ChatGPT names you when a founder asks who to hire. That decision runs on whether your company has said something specific enough for a machine to repeat with confidence.

This isn't a knock on SEO shops. The good ones do real work and the work matters. The problem is the assignment. You handed an execution team a question that gets answered upstream of execution, and everyone has been polite about it for two quarters while the retainer renews.

What can an SEO agency actually move on AI visibility?

More than the AEO tool crowd admits. Four things, specifically, and all four are real.

1. 1Crawlability and rendering. If your product pages are client-rendered, gated, or buried three clicks deep, models never see them. That's a plumbing problem and your SEO team can fix it this month.
2. 2Machine-readable structure. Headings phrased as the questions buyers ask, clean canonical URLs, and markup that tells a parser what kind of page it's reading. Whether schema alone earns citations is a narrower question than most vendors admit.
3. 3Entity consistency. The same company name, the same one-line description, the same claims across your site, LinkedIn, Crunchbase, and every directory you've ever filled out. Engines resolve entities before they recommend them, and contradictions cost you.
4. 4Third-party placement. Engines quote lists far more often than they quote your homepage, so getting onto the credible ones is legitimate, useful work.

Every one of those is inside a competent SEO agency's skill set. If yours hasn't done them, you have a performance conversation on your hands, and it's a fair one to have. Read the detail on what schema markup does and doesn't earn you and on why engines quote third-party lists instead of your site before you grade them.

Where does the SEO retainer hit its ceiling?

At the sentence. Those four jobs make you findable and readable. None of them make you repeatable. When an engine answers "who should we hire to fix our positioning," it isn't ranking pages against each other. It's assembling a recommendation out of the things it can say about you without hedging. If the most specific claim on your site is that you're a trusted partner delivering scalable solutions for modern enterprises, the model has nothing to load into that sentence, and it moves on to a company that gave it something.

This is the gap founders keep hitting when they rank first on Google and never get cited in AI search. Ranking proves a crawler found you. Citation proves an engine understood you well enough to put its own credibility behind the recommendation. Our 19-signal read on what engines actually look for puts narrative clarity above every technical signal on the list, and the order matters.

Why do we keep getting sold SEO for this problem?

Because that's who owns the room. We track a standing set of buyer prompts across ChatGPT, Claude, Perplexity, and Gemini. On the AI visibility topic specifically, four consecutive reads covering roughly 184 answers named zero consultancies. Not us, and not the twelve competing firms we track alongside us. In that entire series, no consultancy brand has ever registered a single mention on that topic. The answers resolve to tracking tools and unbranded SEO content instead.

We're in that zero, by the way. Four articles we published on AI visibility in August have never been named once in those answers. We're telling you that because the number is a map of who's publishing on the subject, not a map of who can help you with it. The AI visibility conversation was built by companies that sell crawling, tracking, and content volume, so the answer a founder gets back is crawling, tracking, and content volume. Then the dashboard shows a score that goes up while nothing in the pipeline moves, which is its own problem with how those tools get read.

What actually decides whether an engine recommends us?

Whether your company has taken a position a machine can restate. An engine recommending a vendor is doing what a buyer does, faster: deciding who it can describe in one specific sentence without getting it wrong. That sentence needs a who, a problem, and a point of view. Most B2B sites supply a category and a tone.

We call the load-bearing layer narrative identity, and it sits underneath the logo, the palette, and the homepage copy. It answers three things: why should a buyer change, why now, and why with you. A site that answers those gives every model a quotable spine. A site that lists features and calls them outcomes hands the model the industry average, which is exactly what solution-centric marketing produces.

“An answer engine can only recommend a company that has already said something worth repeating. Volume is no longer the moat. Perspective is.”

How do we tell which layer is broken?

Run two tests this afternoon. First, open ChatGPT and Claude and ask what your company does and who it's for. If both describe you accurately, retrieval is working and your SEO agency has already done its job. If they describe you well and still never recommend you, you've found the gap between being understood and being chosen. Second, run the Cover-the-Logo Test on your own homepage: hide the logo, show it to someone outside your industry, and ask who this is for. If they can't tell, no amount of technical work will make an engine sure of it either.

Four signals say the problem sits upstream of the retainer:

1. 1Models describe you correctly but recommend competitors when a buyer asks who to hire.
2. 2Your three closest competitors' homepages could swap headlines with yours and nobody would notice.
3. 3Reps tell three different versions of the story and each one works, which means none of them is the story.
4. 4Your visibility dashboard is trending up and your booked calls are flat.

What should we ask our SEO agency on the next call?

Three questions separate the work they can do from the work they can't, and none of them are a trap.

1. 1Which of our pages can a model read end to end, and can you show us the ones it can't? A good agency has that answer ready, and the fix is technical and fast.
2. 2When you write for us, where does the claim come from? If the honest answer is a competitor scan, a keyword tool, or the previous version of the site, you've found the vacuum. Nobody has handed them a position to write from, so they're reconstructing one from the category.
3. 3What would you need from us to make your work land harder? The good ones all say a version of the same thing: a real customer definition, a clear point of view, and proof we can point at.

Notice that none of those get answered by a dashboard. They get answered by what your leadership team has decided, or hasn't. Engines resolve you the same way a skeptical buyer does, which is why entity authority and narrative clarity move together rather than as separate projects.

Do we fire the SEO agency?

Probably not. Change the input and keep the executor. Agencies are execution engines, and they fill whatever the brief leaves empty with the industry average, which is the mechanism behind every agency that has failed you so far. Hand the same team a documented position, the disqualifiers, the villain you're fighting, and the proof, and their technical work starts compounding instead of amplifying a blur.

The sequence is the whole answer. Decide what you stand for, document it so humans and machines read the same thing, then spend on retrieval. Run it the other way and you've funded your own invisibility, which is the same trap founders hit when they weigh the next marketing dollar between AEO and SEO without fixing what either channel will carry.