Presence
Visibility rate, share of model, answer inclusion and the competitive co-mention set. Whether you enter consideration at all.
Answer Engine Optimization
Be the source AI answers recommend
Not the ad beside them. Answer Engine Optimization makes your expertise the thing an AI platform retrieves, cites and recommends.
The click is disappearing
Your buyers increasingly describe their situation to ChatGPT, Gemini or Perplexity before they ever type a query into a search box — and they act on the answer they get back. That answer is assembled from sources. Either you are one of them, or a competitor is.
Paid click-through on those same AI Overview pages fell from 8.76% to 6.56% over the same period. This is not a reason to abandon search — paid search still captures buyers who have already declared intent. It is a reason to compete earlier, in the research and shortlisting that now happens inside an answer engine, before intent becomes visible in any conventional media signal.
The distinction that matters
Mention, citation, recommendation
-
01
Mention
Your brand is named somewhere in the answer. Pleasant, and largely inert.
-
02
Citation
One of your sources is explicitly referenced as supporting the answer — a link, footnote or source card.
-
03
Recommendation
The model actively suggests you as the answer. This is the one that moves a buyer, and it drives 2.5× the referral traffic of a neutral mention.
Most AI visibility reporting counts mentions. Counting mentions tells you almost nothing about whether anyone acted.
Recommendation Stance
We track whether AI platforms recommend your brand, present it neutrally, or steer the buyer to a competitor — prompt by prompt, across the questions that matter commercially.
Visibility tells you that you appeared. Stance tells you what appearing did. A brand can be highly visible in AI answers and still be losing, because every answer frames it as the expensive option, or the one that doesn’t serve your buyer’s segment. That is a solvable problem, but only once you can see it.
What we measure
Five families, not a single score
Prominence
Recommendation position, depth of description and first-source prominence. Separates a passing mention from a meaningful recommendation.
Authority
Citation frequency and share, source domains and third-party authority. Shows which sources actually cause an engine to trust and retrieve you.
Accuracy & safety
Factual accuracy, stale claims and hallucination severity. Stops visibility gains from turning into misinformation about your own products.
Sentiment & association
Positive, neutral or negative framing, and which attributes attach to you. Measures whether visibility reinforces the positioning you intended.
Prompts are not keywords
You cannot reuse an SEO keyword list for this, and most of the AI visibility tooling on the market quietly tries to. Keywords are short and stripped of context. Prompts are long, qualifier-rich and situational — people tell an answer engine things they would never type into a search box.
-
Manufacturing
“What should a mid-market manufacturer measure instead of cost per lead when the sales cycle runs nine months?”
-
Financial services
“What should a mid-sized advisory firm look for in a client portal before its next compliance review?”
-
Enterprise software
“Which platforms can a 500-person company roll out without an in-house data team?”
-
Professional services
“How do consulting firms usually price a six-month transformation engagement?”
-
Telecom
“What should a multi-site business weigh up before replacing its network provider?”
-
Healthcare
“Which systems suit an independent specialty practice with fewer than fifty providers?”
We build a versioned prompt panel across your priority segments, products and buying moments, written at several levels of intent — education, problem recognition, comparison, eligibility and purchase. Every prompt is versioned and re-tested over time, because the answers change even when the question doesn’t.
Coverage
Where we can see — and where nobody can yet
Tracked today
- ChatGPT
- Google Gemini
- Perplexity
- Llama (Meta)
- Google AI Overviews
- Claude
Not yet measurable
- TikTok
- Amazon (Nova)
- Brand-owned chatbots
We publish the second list because the gaps matter. If a channel in your paid mix sits outside AI visibility measurement, you should know that before you plan against it.
Built on content and measurement you already need
Answer Engine Optimization is a distinct capability, but it is not a separate discipline bolted onto the side. It runs on two things we already do.
The Content blueprint produces what an answer engine can actually parse and corroborate: clear explanations, comparison logic, definitions, structured data and visible update dates, built inside your approved legal framework. The Optimization blueprint supplies the closed-loop measurement that connects AI visibility to branded demand, qualified visits and revenue — rather than leaving it as a vanity score in a separate dashboard.
That is the difference between an agency that reports on AI visibility and one that can move it.
60–200%
average improvement in LLM visibility
Horizon internal research
What we’re honest about
The limits of this discipline, today
AI discoverability is a new measurement practice. Overclaiming precision would cost more credibility than it buys, so these are the constraints we plan around rather than paper over.
- No year-over-year benchmarks. The discipline is too new. Tracking begins the day the work starts, not before.
- Visibility scores are directional. Useful for comparison against competitors over time; not decision-grade on their own.
- There is no equivalent of search volume. No platform reliably reports how often a prompt is actually asked.
- Answers are probabilistic. The same prompt can return different answers, which is why panels are versioned and re-tested rather than sampled once.
Common questions
What is answer engine optimization?
Answer engine optimization (AEO) is the practice of making a brand the source that AI answer engines — ChatGPT, Gemini, Perplexity, Google AI Overviews and others — retrieve, cite and recommend when someone asks a question. Where traditional SEO competes for a ranked link a person clicks, AEO competes to be inside the synthesised answer itself.
How is AEO different from SEO?
SEO optimises a page to rank in a list of results. AEO optimises a brand’s wider evidence base — owned content, structured data and third-party citations — so an answer engine trusts and recommends it. The two overlap, but they are measured differently: SEO counts rankings and clicks, while AEO counts presence, citations and whether the model actually recommends you.
What is generative engine optimization (GEO)?
Generative engine optimization, or GEO, is used more or less interchangeably with AEO. Both describe optimising for AI-generated answers rather than ranked links. The terminology is still settling; you will also see AI search optimization, LLM visibility and AI discoverability used for the same work.
Can you measure whether AI recommends my brand?
Yes. We track Recommendation Stance — the share of relevant AI responses that recommend your brand, present it neutrally, or favour a competitor — across a versioned panel of prompts, tested repeatedly across the major platforms. It is distinct from simple visibility, which only tells you whether you appeared.
Does AI search replace paid search?
No. Paid search remains essential when a buyer has already declared intent. AI search creates an additional opportunity to influence research, evaluation and shortlisting that happens before that intent is visible in any conventional media signal. The two are complementary, and the budget question is how to balance them.
How do you attribute business outcomes to AI search?
Carefully, and with a hierarchy of evidence rather than a single last-click number. Many answer journeys are zero-click and referrals are often stripped, so we combine tagged referrals, time-aligned changes in branded demand, staged experiments across content clusters, and — once there is enough history — modelled contribution inside marketing mix modelling.
Find out how AI answers describe you today
We’ll run a sample of real buyer prompts across the major answer engines and show you where you appear, how you’re characterised, and whether you’re being recommended.
Book a 30-minute consultation30 minutes · Virtual · No obligation