What SEO Teams Should Stop Doing First

SEO teams are the best-prepared people in any organisation for AEO, and they are also the most likely to spend their first quarter on the wrong things. Both facts have the same cause: two decades of well-drilled habits, most of which transfer and a few of which quietly do not.

Almost every guide written for this audience is additive — here are the new tactics to layer on. That framing is comfortable and it is the reason so many programmes stall, because the constraint is rarely a missing tactic. It is a familiar routine consuming the hours the new work needs.

So here is the subtractive version: what to stop, what to keep, and how to tell the difference.

Stop reporting position as a single number

Average position is the most load-bearing habit to break, and it breaks for a structural reason rather than a measurement one.

A ranking is a position in an ordered list of ten links. An AI answer is prose that names two or three brands. There is no eleventh result to be at, no page two, no “we moved from 8 to 6.” You are either in the answer or you are not, and if you are in it, how you appear matters more than the order — a first mention with no reasoning attached is worth less than a third mention that says why you fit.

The replacement is not one metric but a small set: presence, position quality, and share against competitors. Mentioned is not recommended covers the taxonomy. The habit to drop is the instinct to compress it back into a single rank-like figure, because the compression is what destroys the information.

Stop treating a keyword as the unit of work

Keyword-level thinking survives surprisingly well right up to the point where it fails completely.

The failure is query fan-out: a conversational surface decomposes one user question into several background searches and synthesises across the results. You are not competing for the question the user typed; you are competing across its components. A single excellent page can lose to a competitor’s five adequate ones, because the competitor appears in four sub-searches and you appear in one.

That inverts a core SEO instinct — depth on one asset beats thin coverage — and it is worth being precise about the inversion. Depth still wins within a sub-question. Breadth wins across the decision. Query fan-out and what Google AI Mode changes works through it.

Stop optimising for a search volume that does not exist here

Every prioritisation ritual an SEO team owns runs on volume, and there is no volume figure for AI prompts. No public query log, no scrapable surface, no vendor with access. Anything presented as prompt volume is a panel, a search-volume proxy, or a model-generated estimate.

The habit to drop is sorting a backlog by a number. The replacement is sorting by commercial proximity — how close is this question to a purchase — using Search Console queries, sales calls and support tickets as the demand signal. That is a worse-feeling process and a better one, since volume always over-weighted the top of the funnel where it is cheapest to be irrelevant. There is no keyword volume for AI prompts covers the substitutes.

Stop reading week-over-week movement

SEO trained everyone to watch rankings closely, because rankings are comparatively stable and a real move means something.

AI answers are non-deterministic. The same prompt run twice returns different brands with nothing having changed. A mention rate computed over a hundred observations carries roughly a ten-point margin, so “we went from 40% to 45%” is entirely consistent with nothing having happened. Reporting that as improvement is not a rounding error; it is attributing a random draw to last month’s work, and it leads directly to scaling a tactic that did nothing.

Read trends across several scans, not deltas between two. How many runs before you trust an AI visibility number has the arithmetic.

Stop auditing for the shadows of quality

The E-E-A-T checklist industry — author boxes, credentials blocks, “reviewed by” badges — was always optimising for the observable correlates of expertise rather than expertise itself. E-E-A-T is a framework in Google’s search quality rater guidelines, written for human evaluators. It has never been a ranking signal, and there is nothing to raise.

In an AI context the error gets more expensive rather than less, because a model composing an answer is even further from parsing a trust badge than a ranking system was. It is reading your claims and deciding whether they are usable. E-E-A-T in the age of answer engines separates what transfers from what never worked.

What carries over unchanged

The subtractive list is short because most of the discipline holds. Four things transfer with no modification, and they are the four SEO teams are already best at.

Technical accessibility. If a crawler cannot fetch or render your content, nothing else matters. Server-side rendering, sane status codes, a robots.txt that does not accidentally exclude the search crawlers — this is exactly the same job, against a slightly different fleet of bots. Should you block AI crawlers covers the fleet.

Information architecture. Retrieval operates on passages, not documents. A page whose sections are self-contained and clearly delimited can be retrieved for a specific sub-question; one woven into a continuous narrative cannot. Good IA was always this; it now has a sharper payoff.

Authority through third parties. The mechanism generalises from links to consistent corroboration across reviews, documentation, press and community discussion. Links are one form of that. The instinct — that what others say about you outweighs what you say about yourself — is exactly right.

Structured data, for entity resolution rather than rich results. Organization markup with sameAs is the clearest disambiguation signal available, and entity confusion is a leading cause of engines stating wrong facts about a brand. Does schema markup help AI visibility is careful about which claims survive scrutiny.

The one genuinely new skill

Everything above is a keep or a drop. One thing is an addition, and it is the least SEO-shaped work on the list: reading answers.

There is no equivalent activity in search. You did not read the SERP to understand why you ranked; you read a rank-tracker. In AEO the answer text is the diagnostic. Whether you were named, whether the framing was favourable, which competitors appeared, what sources the engine reached for — none of it is in a number, and all of it is in the prose.

The teams that move fastest are the ones who make reading a fixed weekly habit for twenty prompts, and the ones that stall are the ones that build a dashboard and never open the underlying answers. That is a genuine behaviour change, and it is the one worth protecting the hours for.

The counter-argument

The fair objection: this reads as “your expertise is partly obsolete,” delivered to the people whose expertise is most relevant, and teams have heard that before about voice search and mobile-first and it was mostly overblown.

That scepticism is well-earned and it should temper the framing. Nothing here says SEO is over — the transfer list is longer than the stop list, and the technical half of the discipline is not merely relevant but load-bearing. The claim is narrower: five specific habits are calibrated to a system with different properties, and continuing them produces confident readings of a signal that is not there.

There is also a real risk in the other direction, worth naming. A team that abandons rank tracking entirely because AI is the future will discover that Google still sends most of their traffic. Running both is the correct posture, and do I need AEO if I rank on Google is the honest version of that trade-off.

Where to start

1. Take one hour and read twenty answers for prompts that matter commercially. Not a dashboard — the raw text. This will reorder your priorities more than any audit.
2. Split your reporting so AI presence sits beside organic performance rather than inside it. Merging them corrupts both.
3. Keep doing the technical work. It is the same work, and it is the most common single cause of invisibility.

AEO for SEO professionals covers how the two practices sit together in a team, and AEO vs SEO is the conceptual comparison if you want the theory before the reshuffle.

The summary a rank tracker cannot give you: you are further along than most functions in your company, and the last mile is unlearning five habits rather than learning fifty tactics.