Fundamentals
"AEO is just SEO." Nine objections from SEO consultants, answered with data.
By Arnav Mukherjee, founder of TofuBofu · August 11, 2026
The most useful pushback I have had on this product came from people who have done search for fifteen years. They have watched three or four categories get renamed and resold, they can smell a repackaged deliverable, and their instinct that most of this is old work wearing a new label is correct more often than the vendors in my category will admit.
So this is written for that reader, and it concedes where conceding is honest. Three of the nine objections below are right, and one of them is right about something my own industry routinely oversells. The other six are answered with numbers we measured ourselves rather than with a narrative about how everything has changed.
The position I will defend is narrow and it has not moved: search rank is the floor, and being named in an answer is a distinct layer built on top of it that you can win even when your ranking does not move.
The six where the data does the arguing
1. "It is the same work. Good content, good technicals, good links."
The inputs overlap heavily and the outcomes come apart, which is the whole argument in one line. Our regional research checked every firm the engines named against those regions' own directories: 139 of 228 directory-listed firms were named by no engine at all. The control group is sharper still. Of 88 firms ranked on a published national industry list, 66 were named by nobody. These are findable, credible, frequently well-ranked companies. Whatever the engines were selecting on, presence and quality alone did not deliver it.
2. "If you rank number one you will get cited anyway."
Often true, and not dependable enough to be a strategy. Retrieval leans on the same index, which is exactly why we say rank is the floor rather than a distraction. But an engine retrieves several sources and then decides whose name to actually say, and that second step is where corroboration across independent sources does the work. In practice we keep finding firms that rank organically in their market and appear in no engine's answer for their own category question.
3. "Schema is not a ranking factor and it never was."
Correct, and our own data is the best evidence against people overselling it. Across 14 matched pairs, Organization markup sat on 86 percent of the firms no engine named and 88 percent of the firms it did. More of the invisible firms served an llms.txt than the named ones. Schema is a wrapper that makes existing content machine-readable. It cannot wrap an answer nobody wrote, and any vendor whose pitch peaks at structured data is selling plumbing as strategy.
4. "This is just brand building with a new invoice."
The closest to right of any objection here, and I would rather agree than dodge. The mechanism genuinely is corroboration, which is brand work by another name. The difference worth paying for is the audience: a retrieval system with no memory of your advertising and no access to your reputation unless it was published. That is why a firm can be famous in its industry and absent from the answer, and why classic brand tracking will not surface the problem.
5. "AI referrals are one percent of sessions. The maths does not work."
The session count is real and it is the wrong denominator. The decision happens inside the answer, where a buyer narrows to three names and never clicks anything to do it, so referral traffic measures the leftovers rather than the event. G2's 2026 research puts 51 percent of B2B buyers starting vendor research on an AI chatbot, up from 29 percent, with 69 percent having switched vendor based on what AI told them. Judge the channel on whether you are in the shortlist, not on sessions.
6. "Google will absorb all of this and it will just be search again."
Partly happening, and it cuts the other way. AI Mode data now lands in the main Search Console performance report rather than a separate generative one, which is consolidation on Google's surfaces. ChatGPT, Claude and Perplexity are not Google properties, report to nothing, and disagree with each other constantly in our measurements. A practice that only watches Google is watching one of several rooms where the shortlist gets built.
Where the two practices overlap, and where they come apart
Settle it on a client, not in an argument
Run a free scan across six AI engines on a client's real buying questions and compare it to where they rank. Ten minutes, and the gap is either there or it is not.
Get your free auditThe three where the consultant is right
These are the ones my own category should stop arguing with. Each is a real limitation, and pretending otherwise is why the objection exists in the first place.
7. "The measurement is not stable enough to bill against."
Right. Answers are non-deterministic, so the same question asked twice can name different firms. Across five US metros we found 85 percent of firms named by exactly one of the four answering engines and none named by all four. The fix is not a better single number, it is sampling each question several times per engine, reporting per engine rather than blending, and reading a trend rather than a snapshot. Anyone selling you one confident score off one run is hiding the variance rather than handling it.
8. "Nobody can attribute an AI mention to revenue."
Also right, and the honest position is to say so. ChatGPT and Perplexity pass referrer information; several others do not, so their referrals land in direct traffic. A mention that shapes a shortlist and produces a search for your name weeks later cannot be traced at all. What you can do is measure the leading indicator, which is whether you appear on the questions your buyers ask, and ask new leads how they found you. Any vendor promising clean revenue attribution here is selling something nobody can currently deliver.
9. "Half the numbers in this category are made up."
Right often enough that the skepticism is well earned. We have caught bad figures twice in our own checking, including a competitor's published number that was wrong at the vendor's own page, and we shipped a stale statistic in our own FAQ markup and had to correct it. The defence is boring and it works: verify at the primary source, date the figure, and publish the sample size and selection method so the reader can discount you appropriately.
What this means if you run an SEO practice
The practical read is not that your skills are obsolete. It is that your existing work covers the floor well and stops short of the layer, and the layer is where an increasing share of the shortlist decision now happens. Most of what closes the gap is work you already know how to do: get independent sources saying consistent, specific things about the client, and make sure the client's own pages answer the buying question rather than describing the company.
What is genuinely new is the measurement, and it is new in an awkward way. It is per engine, non-deterministic, needs sampling, and refuses to collapse into one number you can put on a slide. That is a real operational cost and the main reason to use a tool rather than a spreadsheet, and it is also the reason to be suspicious of tools that hand you a single confident score.
The reason to care now rather than in two years is client-side. Forrester's 2026 study found 94 percent of B2B buyers use AI somewhere in the buying process. When a client asks why enquiries are drifting down while rankings hold, a practice that measures only the floor has no answer, and the absence of an answer is what gets retainers cancelled.
Frequently asked questions
Is AEO just SEO with a new name?
They overlap on inputs and separate on outcomes, and the separation is measurable. Ranking well is largely necessary and clearly not sufficient: our regional research found 139 of 228 directory-listed firms named by no AI engine at all, and in a control group drawn from a national industry ranking, 66 of 88 listed firms were named by nobody. Those firms are findable. Many rank. The output of an answer engine is a shortlist of names rather than an ordered list of links, and a page can be retrievable without ever being the thing an engine chooses to name.
If I rank number one, will AI cite me anyway?
Often, and not reliably enough to plan around. Rank is a strong input because retrieval leans on the same index, which is why the honest framing is that search rank is the floor. But engines assemble an answer from several sources and then choose whose name to say, and that choice is influenced by corroboration across independent sources rather than by your position alone. The measurable version of the gap is that plenty of firms ranking organically in our study were named by no engine, so a top position did not carry them into the answer.
Is structured data actually a ranking factor for AI answers?
Not in the way it gets sold, and our own data is the strongest argument against overselling it. We crawled 14 firms that AI engines never named against the specific rivals those engines did name in the same regions. Organization markup was present on 86 percent of the invisible firms and 88 percent of the named ones, and more of the invisible firms served an llms.txt than the named ones did. Schema makes existing content machine-readable, which is worth doing. On this evidence it does not separate a cited firm from an uncited one, and any vendor leading with it is selling you plumbing as strategy.
Nobody can measure AI visibility reliably, so why pay for it?
This objection is largely correct and deserves a straight answer. Answers are non-deterministic, so a single check is close to worthless and any honest measurement samples repeatedly and reports per engine rather than blending. Across five US metros we found 85 percent of firms were named by exactly one of the four answering engines and none appeared on every list, which is exactly the instability the objection points at. The right conclusion is not that measurement is impossible, it is that a single blended score is the wrong instrument and a per-engine, per-question, sampled trend line is the right one.
AI referral traffic is about one percent. Why would anyone invest?
Because clicks are the wrong denominator for this channel. The influence happens inside the answer, where the buyer forms a shortlist and never visits a site to do it, so measuring it by referral sessions counts only the residue. G2's 2026 research found 51 percent of B2B buyers now begin vendor research on an AI chatbot, up from 29 percent, and 69 percent have switched vendor based on what AI told them. The honest caveat is that attribution here is genuinely weak, and a vendor claiming clean revenue attribution from AI mentions is overstating what anyone can currently prove.
Is not this just brand building with extra steps?
It is closer to brand building than to technical SEO, and that is a fair reframing rather than a rebuttal. The mechanism is corroboration: several independent sources saying a consistent, specific thing about you. Where it differs from classic brand work is that the audience is a retrieval system with no memory of your advertising, so unpublished reputation counts for nothing. A firm can be genuinely famous inside its industry and absent from the answer, which is a failure mode brand measurement would not flag.
Will not Google just fold all of this back into normal search?
Partly, and it is already happening, which strengthens rather than weakens the argument. AI Mode data now appears in the main Search Console performance report rather than in a separate generative report. But the consolidation is on Google's surfaces only. ChatGPT, Claude and Perplexity are not Google properties and do not report to Search Console, and our measurements show them disagreeing with each other constantly. A practice that only watches Google is watching one of several places where the shortlist gets made.
Sources and further reading
- TofuBofu MSP research 2026: 139 of 228 directory-listed firms named by no engine, and 66 of 88 firms from a national industry ranking likewise.
- Engine disagreement across five US metros: 85 percent of firms named by exactly one of four engines, none by all four.
- Matched-pair crawl of 14 invisible firms: Organization markup on 86 percent of them against 88 percent of the firms the engines named.
- G2 2026 B2B buyer research: 51 percent begin vendor research on an AI chatbot, up from 29 percent, and 69 percent switched vendor based on AI.
- Forrester B2B Buying Study 2026: 94 percent of buyers use AI somewhere in the buying process.