I spend a lot of time in keyword research tools. Not just looking for traffic opportunities — looking for vocabulary signals. The specific words people use to describe a problem or a discipline tell you something important about where a market is in its maturity cycle. And right now, the data is telling an unusually clear story.
Let me show you what I mean. These are real figures from Google Keyword Planner, UK, March 2025 to February 2026.
GEO agency: 210 searches per month, +1,200% year-on-year.
AI visibility: 90 searches per month, +2,000% year-on-year.
AI optimisation: 480 searches per month, +86% year-on-year.
SEO for AI search: 70 searches per month, +800% year-on-year.
And then a second tier — terms with zero measurable volume but that exist in published, authoritative content right now:
AAO agency, AAO consultant, assistive agent optimisation, agentic AI optimisation: all showing as unavailable in keyword tools. Zero volume. No data.
Most keyword researchers see that second tier and skip it. Nothing there. Move on.
I see it differently. Zero volume on a term that was published in Search Engine Land three weeks ago is not a signal to ignore. It is a window.
What Keyword Data Actually Tells You
Keyword volume is a lagging indicator. By the time a term has meaningful search volume, somebody already owns it. The brands that built organic authority on ‘inbound marketing’ in 2008, ‘growth hacking’ in 2013, ‘content marketing’ in 2014, ‘entity SEO’ in 2019 — they did not do it by finding high-volume terms. They did it by being early to a vocabulary that was still forming.
What the data is showing right now is a vocabulary in formation. The market is trying to name something — the discipline of making your business visible in AI-powered discovery systems — and it has not settled on the language yet. GEO, AIO, AEO, AAO, AI SEO, LLM optimisation: these are all candidates. The winner will be the term that gets attached to the most authoritative, most cited, most structured content. That term will own the search results when the volume arrives. And the volume will arrive.
This is not speculation. This is how every new search discipline has developed. The pattern is consistent: concept emerges in practitioner writing → vocabulary fragments → one or two terms gain traction → volume concentrates → late movers try to buy their way in and mostly fail.
We are currently between step two and step three. The window to establish authority before volume concentrates is open. It will not stay open long.
The Three Trends the Data Is Capturing Simultaneously
What makes this keyword dataset unusual is that it is not capturing one trend. It is capturing three markets merging.
The first is AI visibility services — the emerging commercial market for helping businesses appear in AI-generated search answers. GEO agency, AIO agency, AI visibility tools, AI citation. This is the direct commercial expression of the shift. Businesses are searching for it because they know something has changed and they need help with it.
The second is traditional SEO in transition — terms like SEO for AI search, SEO strategy agency, technical SEO agency showing large YoY percentage increases (SEO for AI search at +800%) alongside declining absolute volumes for some traditional terms. The market is not abandoning SEO — it is demanding that SEO practitioners understand AI. The agencies that position at this intersection, rather than on one side or the other, are in the strongest commercial position.
The third is the agentic AI ecosystem — and this is the forward-looking signal most people are missing. Agentic AI at 22,200 searches per month, +50% YoY. What is agentic AI at 5,400 per month, +175%. These are not AI SEO terms. They are general technology and business terms. But they are directly adjacent to what is about to happen in search, and the businesses that understand this adjacency before their competitors do will have a significant positioning advantage.
The Piece Published Three Weeks Ago That Changed How I Think About This
In late February 2026, Jason Barnard published a piece on Search Engine Land introducing the concept of AAO — Assistive Agent Optimisation. The argument was precise: every competing acronym (GEO, AIO, AEO, LLM optimisation) covers part of the AI visibility problem, but none covers all of it. AAO is the framework that does, because it names the purpose across the full lifecycle — from search engine to AI answer engine to AI agent making autonomous decisions on behalf of a user.
The piece introduced a concept he calls the Algorithmic Trinity: the three components that every AI discovery system runs on simultaneously. Large language models. Knowledge graphs. Traditional search. The balance differs by platform — ChatGPT leans LLM-heavy, Google leans on its knowledge graph, Perplexity weights its own retrieval index — but all three are present in every platform, all the time.
This framing crystallised something I had been working around for months without quite naming it. The reason so many AI visibility strategies produce patchy, platform-inconsistent results is that they optimise one component of the trinity while leaving the others unaddressed. You can rewrite all your content for AI citation readiness and still not appear in ChatGPT answers because your Bing indexing has gaps — that is retrieval failure, not selection failure, and they require completely different fixes. You can fix your technical infrastructure and still not be named as a recommended provider by an AI system because your entity authority is insufficient — that is Layer 4 failure in a different model, and content work does not fix it.
The Framework That Emerged From All of This
Taking the keyword data, the Algorithmic Trinity, the AAO concept, and two decades of practitioner experience, I built what I am calling the AI Discovery Stack — a five-layer model that maps how AI systems actually progress from finding a website to citing it or acting on it autonomously.
Layer 1 — Understanding. The entity layer. The AI system asks: who are you, what are you, are you credible? Schema markup, knowledge graph signals, author identity, cross-platform entity consistency. Without this, AI systems cite you inconsistently or not at all — not because your content is poor, but because they cannot confidently identify what you are.
Layer 2 — Retrieval. The traditional SEO layer. Can AI systems find and index your content? This is where Bing indexing matters — Bing feeds ChatGPT Search and Microsoft Copilot, so a page absent from Bing simply does not exist for those platforms. Page speed, crawlability, robots.txt configuration. Classic technical SEO, but with AI crawler requirements that are in some respects more demanding than Googlebot.
Layer 3 — Selection. This is where AI diverges most sharply from traditional search. A page can be perfectly indexed and authoritative and still not be selected as a citation source if its paragraphs are not structured for extraction. AI systems select at the paragraph level, not the page level. They prefer standalone answer openings, explicit term definitions, attributed statistics, and heading structures that map directly to the sub-questions being asked. The Princeton GEO-Bench study found that specific structural techniques improve citation visibility by 30–40% across 10,000 AI-generated responses. That is not a small effect size.
Layer 4 — Recommendation. There is a meaningful difference between being used as an anonymous source in an AI answer and being named as a recommended provider. Layer 4 is the difference. Entity prominence, external citation networks, digital PR, and the richness of your knowledge graph representation determine whether the AI says ‘according to sources’ or ‘according to SEO Strategy Ltd.’
Layer 5 — Action. The agentic layer. AI does not show a list and ask the user to choose — it evaluates, selects, and acts. The entire buying funnel happens inside the agent before the human sees a result. Being selected at Layer 5 requires strength at Layers 1 through 4 as prerequisites. AAO is not a separate discipline. It is the terminal layer of the same stack.
The practical value of the model is diagnostic. When an AI visibility strategy is not working, the question is not ‘what should we do?’ — it is ‘which layer is failing?’ Different layers have different symptoms and completely different fixes. Applying content structure improvements to a retrieval failure wastes budget. Fixing technical indexing when the real problem is selection failure changes nothing.
The Thesis Underneath It All
Here is the thing I keep coming back to, the thing that the keyword data, the AAO concept, and the stack all converge on.
Visibility has always been the goal. The landscape is just bigger now.
SEO gave you one search engine. AEO gave you voice and AI answers. GEO gave you generative responses. AIO gave you AI Overviews. AAO gives you agentic decisions happening without a human ever seeing a search result. The discipline — identifying where your audience finds information and making sure you are the answer they find — has not changed. The execution has.
The businesses that understand this continuity, and act on it while the keyword curve is still forming, will be the ones that dominate AI-era visibility the way early SEO adopters dominated Google. The ones that wait for the volume to arrive will be late — buying authority that the early movers built for free.
What to Do With This
If you are a business owner or marketing director reading this, the immediate question is whether your current SEO strategy addresses all five layers or just one or two. The most common failure mode right now is strong traditional SEO (Layer 2) with no entity architecture (Layer 1) and content that is not structured for AI extraction (Layer 3). That combination produces good Google rankings and poor AI visibility — which was fine eighteen months ago and is increasingly problematic now.
Start by running your key queries through ChatGPT, Perplexity, and Google AI Overviews. Are you being cited? Are competitors being cited instead? If competitors are appearing and you are not, the gap is almost certainly at Layer 1 or Layer 3 — either your entity is not clearly understood, or your content is not structured for extraction. Both are fixable. Neither is fixed by writing more blog posts.
If you want to understand where specifically your AI visibility is failing, the AI Visibility Audit diagnoses exactly that — across all five layers, with a prioritised remediation plan. Or read the full AI Discovery Stack guide, which explains each layer in detail with worked examples from this site.
The keyword curve is readable right now. Most people are not reading it. That is the opportunity.
