01
Google
AI Overviews
Synthesised above everything, drawing from a handful of sources it considers settled. Hardest to enter, most valuable to hold.
Sampled nightly.
Answer Engine Optimization
Be the answer, not the fourth blue link
The result page stopped being a list. AI Overviews, featured snippets and voice assistants now compose one answer and name one or two brands. AEO is the discipline of making sure the composed answer contains yours.
~60%
of searches now end without a click to any website
1-2
brands typically named inside a single composed answer
412
prompts we sample nightly to see who gets named
What changed underneath you
I
Position one is not the top of the page
Above your ranking sits a synthesised answer assembled from several sources. If you are not one of them, your #1 position is below the fold of the decision.
II
Answers are extracted, not ranked
The engine is looking for a clean, self-contained, attributable passage. Pages written as long persuasive essays give it nothing liftable, no matter how good they are.
III
The question is the unit, not the keyword
Buyers type sentences now. We map the full question space of your category, including the awkward, comparative and objection-shaped ones your marketing avoids, and cover it deliberately.
The Test
Could a model quote one paragraph of this page, out of context, and be correct, specific and unambiguous about who said it?
If the answer is no, the page cannot be cited, only summarised. Summaries do not carry your name.
Self-contained passages
Unambiguous attribution
Answer-first order
The AEO system
Four moves, run in order. Skipping one wastes the rest.
We assemble the real question space from search data, sales-call transcripts, support tickets, Reddit threads and direct prompting of the engines themselves. Then we score each question by commercial value and by how contested the answer currently is. That intersection is your roadmap.
Prompt mining
Sales-call transcripts
Contest scoring
The claim in the first forty words. Definitions that stand alone. Comparison tables an engine can parse. Lists where the list is the answer, not decoration. We restructure existing high-intent pages before writing new ones, the fastest wins are usually already published.
Forty-word answers
Standalone definitions
Parsable tables
Facts placed near stable anchors, unique data the engine cannot source elsewhere, clear author and organisation entities, and schema that removes ambiguity about who is speaking. When the model needs a source, you want to be the least effortful one to name.
Stable anchors
Unique facts
Author entities
Schema
Your question set is sampled across five engines every night. You see which questions name you, which cite you without naming, and which name a competitor, with the actual answer text, not a score. Gaps become the next sprint.
5 engines
Answer text captured
Gap-to-sprint loop
Where answers get composed
Each surface has its own extraction behaviour. We treat them separately because they behave separately.
02
Google
Featured snippets
Still the training ground: if you cannot win a snippet on a question, you rarely make the Overview for it either.
Structure-driven, winnable fast.
03
Perplexity
Cited answers
The most attribution-friendly surface. Numbered sources, visible, clickable, and heavily weighted toward original data.
Our best-performing surface.
04
Voice
Spoken results
One answer, no list, no second place. Requires the shortest, most declarative version of your claim to exist somewhere.
Often overlooked, cheap to win.
The first ninety days
- 01 Baseline reading Your question set sampled across five engines. You see exactly where you stand before we touch anything. Week 1
- 02 Question map The full space, scored by value and contest. Agreement on the first thirty questions to own. Week 2
- 03 Restructure what exists Answer-first rewrites of pages already ranking. The fastest movement in the whole engagement. Week 3-5
- 04 Publish into gaps New pages for uncontested high-value questions, each carrying original information. Week 5-12
- 05 Hold the position Nightly sampling, monthly refresh, and a defence sprint whenever a competitor takes a question back. Ongoing
Next
Find out which answers already name you
Send your ten most valuable buyer questions. We run them across five engines and send back the answer text, who gets named, and what it would take to be you instead.