AEO vs SEO vs GEO: A Claim Check on What the Vendors Say

"Answer engine optimisation is a separate discipline from search engine optimisation." That sentence, or a close variant, opens most vendor guides in this category, and the boundary moves depending on who is drawing it. Before anyone renames a team around it, we traced five of the claims these guides repeat back to whatever is actually underneath them.

For a marketing lead who needs one plan rather than three. We read the primary study where one existed, and we name the ones we could not open. We build content software, which makes us an interested party; no competitor's product is being tested here, only the claims in public pages.

Five claims, five sources, five verdicts

What gets repeated	Where it leads	Where we land	In a sentence
The newer acronyms describe separate jobs	Writesonic, Writer, Jasper, each differently	Partly holds	They point at different surfaces, not different work
Classic search optimisation has been superseded	Google; the vendor guides	Does not hold	Google asks for nothing extra, and citations lean on ranking pages
An AI Overview costs position one 58 per cent of its clicks	Ahrefs, February 2026	Holds, with a date attached	Real study, but an earlier edition said half as much
Assistant traffic converts 4.4 times better	Semrush, cited second-hand	Unverified	We never reached a primary page with a sample
Two thirds of searches finish without a click	Unattributed; SparkToro measured lower	Does not hold as stated	The most recent primary figure is well under that

Claim: the two newer acronyms describe two separate jobs

Source. Every guide defines both terms, and no two definitions agree. Writesonic's comparison page treats answer engine optimisation as making content discoverable and easy for answer engines to surface. Writer's guide assigns AI Overviews and knowledge panels to answer engine optimisation, and puts ChatGPT, Claude and Perplexity under generative engine optimisation. Jasper's version inverts part of that, placing Perplexity's instant answers and Bing Copilot under the answer side.

What the comparison shows. One surface, Perplexity, lands in a different bucket depending on the page you read. Wikipedia's entry on the field says the same thing from outside the vendor set, noting that usage varies across practitioners, vendors and publications with no consensus definition.

Where we land. Partly holds. The labels do gesture at genuinely different surfaces. They do not describe different work, and the task list underneath them is close to identical.

Use. Define the term once, in writing, and list the surfaces you intend to watch. Then never argue about the acronym again.

Claim: classic search optimisation has been superseded

Source. A framing that follows naturally from having a new product to sell.

What the documentation says. Google's page on AI features and your website states that no additional requirements and no special optimisations are needed to appear in AI Overviews or AI Mode, and that a page needs to be indexed and eligible to be shown with a snippet.

What the vendors' own data says. Writesonic's analysis of a million AI Overviews found that 40.58 per cent of citations came from Google's top ten results. Jasper relays an Ahrefs study of 863,000 result pages reporting that close to 40 per cent of AI Overview citations already rank in the top ten, and close to 70 per cent in the top hundred. Both come from interested parties and they measure slightly different things, and they agree on direction anyway.

Where we land. Does not hold. The parties with the strongest commercial reason to claim a replacement are publishing the numbers that argue against one.

Use. Keep the fundamentals funded, add measurement of answers on top. Our guide to writing for answer engines covers the overlap.

Claim: an AI Overview costs the top result 58 per cent of its clicks

Source. This one traces further than we expected, and it has a twin that nobody distinguishes it from. Ryan Law published an Ahrefs study in April 2025 covering 300,000 keywords, half with an AI Overview present and half informational keywords without one, using aggregated Search Console click-through data for March 2024 against March 2025. Its finding was that an AI Overview correlated with a 34.5 per cent lower average click-through rate for the top-ranking page.

What the newer edition reports. Law re-ran it with the same sample design in February 2026, this time comparing December 2023 with December 2025, a two-year window instead of one. That version reports a 58 per cent lower click-through rate for position one. So the figure in the vendor guides is real and recent. It is also one of two published numbers from the same author with the same method and a different window, and almost nobody citing it says which.

Where we land. Holds, as long as the year and the window travel with it. Ahrefs itself frames the relationship as correlation and acknowledges that its forecast assumes the AI Overview keywords would otherwise have followed the control group's trend, which the study does not test. Pew's tracked browsing in March 2025 points the same way from a different angle, with a conventional result clicked in 8 per cent of visits where a summary appeared against 15 per cent where none did.

Use. Cite it as the February 2026 Ahrefs figure over a two-year window, or say clicks fall and cite the study. Never quote the bare number.

Claim: visitors arriving from assistants convert 4.4 times better

Source. Jasper attributes it to Semrush. We did not reach a Semrush page publishing the figure with a sample, so for us it remains a citation of a citation. Separately, AirOps' tracking guide offers a range of two to six times for the same idea, drawn from company self-reports summarised in a webinar recap.

Where we land. Unverified, and the two numbers do not describe the same quantity. A multiplier on conversion rate depends entirely on what the comparison group is, and neither source states one.

Use. Measure your own. Note that many assistants strip the referrer, so a meaningful share of this traffic lands in your reports as direct and your own figure will understate it. Our piece on joining analytics and Search Console data sets up the measurement.

Claim: two thirds of searches finish without a click

Source. Writesonic's page asserts a figure above two thirds and names nobody. The team that actually measures this is SparkToro, working with the Datos clickstream panel, and Rand Fishkin's July 2024 edition reports 58.5 per cent of US searches and 59.7 per cent of EU searches ending with no click, from panel data running September 2022 to May 2024.

Where we land. Does not hold as stated. The most recent primary measurement sits several points below the claim, and higher versions in circulation come either from an older edition of the same work or from a wider definition that counts a trip to Maps or YouTube as a non-click. Note the age: this edition is now over two years old, and we use it because it is the most recent measurement with a published panel and method that we could open.

Use. If you need the number, give the researcher, the year and the market. If you cannot be bothered to carry all three, the sentence is better without it.

The pattern behind the two groups

The claims that survived are about direction: clicks are falling, fundamentals still matter, the acronyms overlap. The claims that collapsed are about precision, and they collapsed the same way every time. A study publishes a figure with a window and a sample. A guide quotes the figure. A second guide quotes the first guide. By the third hop the window is gone, the sample is gone, and what remains is a number that sounds like a law.

Choose the label your finance director already uses

Our position, stated as a recommendation rather than a finding: do not create a role, a budget line or a reporting cadence named after any of these acronyms. Keep the name your organisation already understands, put the new surfaces underneath it as channels, and let the metrics do the differentiating that the vocabulary cannot.

The argument is not that the surfaces are unimportant. It is that the vocabulary has no stable referent, and we have just shown that: the same product, Perplexity, sits in different categories across three vendor pages, and Wikipedia records no agreed definition. A budget line whose definition depends on which guide your new hire read is a budget line that will be re-litigated every quarter.

The outside critique lands harder than ours. Forrester's Nikhil Lai argued in 2025 that these disciplines are "significantly, but not fundamentally, different from SEO" and said the distinction gets inflated to carve out what he described as a startup-sized hole in the marketing stack. Google's own 2026 guidance takes the same line, treating optimisation for generative search as optimisation for search.

The split becomes real on one development only. If a major assistant publishes documentation describing requirements that classic fundamentals do not satisfy, a feed to submit, a markup type that only it reads, an access decision made somewhere other than the web server, then the work genuinely diverges and it deserves its own name and its own owner. Watch for a specification, not for a vendor page. The closest thing to that today is the access layer, and an access decision is a quarter-hour in a dashboard, not a discipline. Where to look first is set out in our seven-blocker audit.

Tracing a number back, step by step

1. Follow the link in the guide, then follow the link on that page, and keep going until you reach something that ran a study rather than quoted one.
2. Write down the window before the result. A one-year comparison and a two-year comparison produced two different headline figures from the same author and the same method.
3. Put two sources on the same claim side by side and write down the gap. The gap is the finding.
4. Check what the publisher sells, and whether a small result would have been published at all.
5. If the trail ends without a method, write "reported by" and the vendor's name, or drop the sentence.

Questions about the three acronyms