Are we optimising for the right engine? Insights from the latest SEO / AEO research

This article is about one of the biggest marketing bubbles (or breakthroughs?) growing right now. AEO or GEO (whatever you call it, it’s basically optimization for AI engines) has become one of the major marketing bets.

First, some facts

Let me start with a bit of context about the new reality we live in – because AI has changed how people make decisions more than you might imagine.

• Fact 1: When an AI Overview is present on Google, users click a traditional result only 8% of the time vs. 15% without one — a 47% relative CTR drop (Pew Research Center, July 2025, based on the analysis of 70K searches)
• Fact 2: Ahrefs found a 58% drop in #1 position organic CTR when an AI Overview is present – analysis of 300K keywords (Ahrefs, Dec 2025)
• Fact 3: AI Overviews now trigger on ~48% of all tracked queries. It's up 58% YoY from 30% a year ago (BrightEdge, Feb 2026)
• Fact 4: Only 17% of AI Overview citations come from pages ranking in the organic top-10 (same study from BrightEdge, February 2026)
• Fact 5: the AI ​​Overview block itself has grown in height from ~1050px to ~1200px over the past year – that's taller than a standard desktop screen (~900px), meaning the top organic result is now physically lower than the first scroll, even if you're #1 in the search results (same study from BrightEdge, February 2026)

Let me translate this: Imagine you’ve spent the last 5 years doing classic SEO. You managed to rank in Google’s top 10 for a number of queries. Now AI is eating up your results, and you’re not even shown in their AI Overview.Yeah, that’s the reality.At the same time, AI is converting way better.

• Fact 1: AI-referred visitors convert at 4.4x the rate of traditional organic visitors (Semrush, June 2025)
• Fact 2: According to a Gartner forecast traditional search engine volume to drop 25% by 2026 as AI chatbots absorb query share (Gartner, February 2024)
• Fact 3: However, only 1% of users clicked a link within an AI Overview itself (Pew Research Center, July 2025)

So users click rarely – but if they do, they are way more likely to convert. Now remember your latest Google search and you're most likely to understand that all these facts are ultra logical and basically mimic your personal behaviour.

Time to optimise for AI then? Not so fast.

Let me start with a bit of a backstory. "GEO" (Generative Engine Optimization) is an academic term coined in a Princeton/Georgia Tech/IIT Delhi/Allen GEO paper (arXiv, Nov 2023), which experimentally demonstrated that certain tactics (citations, statistics, authoritative tone) increase the visibility of content in LLM responses by 37-41% on their benchmark. This is a real, controlled experiment, not marketing hype.

But by 2025-2026, a whole agency fraternity has grown up around this, interested in selling AEO as a separate, revolutionary discipline – and that's precisely why the SEO vs. AEO thesis is far more controversial among practitioners than on agency blogs.

Rand Fishkin (SparkToro) bluntly states that GEO/AEO is simply "SEO for AI," not a new discipline, while Eli Schwartz writes a separate essay, "AEO is Not SEO 2.0," from a more skeptical perspective.

Search Engine Land, Ahrefs, and Semrush have dedicated sections on AI search, but none of the serious publications are burying SEO as a category.

Let's revise the numbers mentioned earlier

In terms of decline figures, the reality is more serious than alarmist headlines, but not in the way predicted. Gartner's famous February 2024 prediction I mentioned of a 25% drop in search query volume by 2026 did NOT come true: Google still holds over 90% of the search market share, partly because servicing a single chat query costs approximately 10 times more than a regular search query, which physically limits the rate of substitution.

However the drop in organic CTR is real. A study by Seer Interactive (Nov 2025, 3,119 queries, 42 organizations, 25.1 million impressions) found a 61% drop in organic CTR for queries with AI Overviews (from 1.76% to 0.61%). That is, it’s not “search is dead,” but “clicks on specific types of queries have dropped sharply in areas where Google itself provides answers directly in the search results.”

Anyways, isn't AEO the same as SEO?

This is a very debatable topic, and I'll try to be more or less objective here. It really depends on the specific platform you're optimizing for, and the difference is huge.

• Ahrefs (Aug 2025, 15,000 queries) found that citations in Google AI Overviews are 76% consistent with the top 10 organic search results –but in their March 2026 report the number declined to 38%. (The BrightEdge report from Feb 2026 mentions a still lower figure – just 17%).
• But for ChatGPT/Gemini/Copilot, the match with Google's top 10 is only about 12%. Perplexity is somewhere in the middle, ~28.6%.

So, the statement "it's the same thing" is arguably true for Google AI Overviews (it's literally an add-on to the same ranking system) – but NOT true for chatbot search, where good positions in traditional SEO have almost no correlation with actual citations. There are other things at work there (that's what they say):

• Citation statistics
• Answer in the first paragraph
• Brand mentions on THIRD-PARTY sites (not your own).

Okay – so AEO is the future?

I wouldn't be so optimistic. Here's a separate, curious and sobering point: Reddit has long been the top source of citations in ChatGPT (about 3% of all citations according to Profound), but according to one Promptwatch, after one of the model updates, the share of citations from Reddit plummeted by 95% in just a week.

The very fact of such volatility is an important lesson: AEO tactics are much less stable over time than classic SEO, where the rules change once every year or two, rather than every time the model is updated.Meanwhile, this market is already very pumped.

• Profound is currently the only confirmed unicorn in the category: Series B at $35M (August 2025, Sequoia) → Series C at $96M (February 24, 2026, Lightspeed, plus Sequoia, Kleiner Perkins, and others) at a $1B valuation.
• Peec AI went from a €7M seed (July 2025) to a Series A at $21M (November 2025, Singular, valuation over $100M with ARR around $4M+) and another round of ~$10M in June 2026 at a valuation of $200M – that is, doubling the valuation in about 8 months with fairly modest revenue.
• Scrunch AI raised a $15M Series A (July 2025, Decibel).
• Athena/AthenaHQ went through YC with a $2.2M seed (June 2025).
• Evertune had a $15M Series A (August 2025).

The biggest move is not a startup, but Adobe: it agreed to buy Semrush for approximately $1.9 billion, and the deal is clearly positioned around GEO/AI search visibility - Semrush reported enterprise ARR growth of 33% year over year. Ahrefs responded by launching its "Brand Radar" AI visibility tracking feature. Optimizely and Conductor did not buy each other separately, but entered into a partnership for a joint AEO platform.So yeah, now GEO/AEO market is clearly hot and in some places overheated capital – unicorn and doubling of the valuation in 8 months with revenue in the millions of ARR – this is a classic pattern of “money chasing the theme faster than real revenue is growing.” At the same time, there are no collapses or closures in the category yet, but most players are less than 18 months old, so the real market test is most likely still to come.

And the final part: Real AEO cases

I tried to find some case studies (including independent experiments) to find if AEO works at all. There is both, and the honest picture is quite sobering: real cases exist, but independent academic verification is much less sweet than these cases.Let's start with what really exists. Profound (an AEO analytics vendor) publishes cases of its clients with names and numbers:

• Ramp made 4 landing pages about AP automation and raised AI visibility from 3.2% to 22.2%, category rank from 19th to 8th place for December 2024 - February 2025 - but the share of citations almost did not move (7.5%→8.1%). The fact that the case honestly shows this is a big "like", vendors usually hide weak indicators.
• 1840 & Co. with one post with FAQ sections increased from 0% to 11% visibility in a month.
• Convert/Omniscient Digital saw an increase in LLM visibility from 31% to 55% in 60 days.
• Novig - from 0% to 21% and 27 times more AI traffic in six months.

Semrush also published its internal experiment (13%→32% per month), but honestly admitted that it could not pinpoint which tactics worked.The problem with all these case studies is the same: they're published either by the vendor selling the AEO tool or by the agency that performed the work – meaning they're a vested interest. The measurement period is short, and there's almost no isolation of cause and effect (anything could have been at play).However, an independent, disinterested review appears far less convincing. A recent (July 2026) academic review of the field – a critical survey on GEO (arXiv, Olivier Martinez) – found only two studies with real traffic/conversion data outside of controlled benchmarks, and both were weak (one with p=0.16 in a placebo test, the other methodologically opaque). Worse still, it cites two independent benchmark studies with negative results:

• C-SEO Bench found a significant positive effect in only 3 out of 54 tactic-domain combinations (and none in the Q&A category),
• SAGEO Arena found that body text optimization actually REDUCES citation frequency by 6-9%.

And that famous "40% increase in visibility" that everyone cites from the Princeton paper I mentioned in the very beginning of this article is outright dismissed by this review as an artifact of a simulator with five pre-posted documents, rather than proof of a real-world increase in discoverability.

The paper itself draws a clear conclusion: claims about GEO ROI clearly outpace the academic evidence. Separately, SEO veteran Lily Ray (March 2026) explicitly warns that several traffic spikes attributed to GEO tactics (scaled content, artificial "freshness," AI-powered resume buttons) actually preceded traffic crashes after subsequent Google ranking updates.

Does that mean that AEO doesn't work?

I will steal this explanation from Lily Ray:

Provide a caption (optional)AEO isn't a "pure experiment without a single case study" – there are real cases with numbers, you can look at them. But even the most rigorous, disinterested testing is currently more likely to be negative or null than confirmatory. I would treat any specific "we did X and grew by Y%" case study as a marketing showcase from one company over a short period, rather than as a reproducible rule – given that independent tests in the same field find either no effect or a negative effect from similar tactics.

I don't have an answer on what you should do next. But we can discuss and share experiments.

So if this feels relevant – I welcome you to my community of growth experimenters on Telegram <3