Be the answer, not the tenth blue link.
Buyers increasingly ask an AI assistant before they ever reach a search results page. Decifer measures whether you are found across all three layers of search, classic, answer-box and AI-generated, and hands you the moves that win the answer back.
Search stopped being ten blue links.
A growing share of searches now end without a click. Google answers at the top of the page; ChatGPT and Perplexity answer before a results page exists at all. The buyer reads one synthesised paragraph that cites three or four sources, and the decision is half-made before any site is visited.
When there is no page two, ranking eleventh is the same as not existing. The unit that matters shifts from share of search, where do I rank, to share of answer, how often am I the source the model trusts.
Then
Ten ranked links
Win by climbing the list. Position three still gets a click.
Now
One answer, a few sources
Win by being cited. Source four is invisible.
Three layers of being found.
They are not the same problem, and a brand can be strong in one and absent in another. Decifer scores all three and tells you which one is costing you the most.
Search Engine Optimisation
Ranking in the ten blue links
The classic surface: Google and Bing returning a ranked list of pages. Won on crawlability, topical relevance, link authority and technical health, indexation, Core Web Vitals, internal linking. Still the largest source of intent, but a shrinking share of it now ends in a click.
What we measure
Rank, impressions, click-through and the queries you own versus the ones you should.
Answer Engine Optimisation
Being the extracted answer
Featured snippets, People Also Ask, voice results, Google lifting one concise answer out of a page and placing it above the links. Won with question-led structure, tight answer blocks near the top of the page, and schema (FAQPage, HowTo) that makes the answer machine-extractable.
What we measure
Snippet ownership, PAA coverage, and how much of your traffic is answered without a click.
Generative Engine Optimisation
Being cited inside the AI answer
ChatGPT, Perplexity, Gemini and Google's AI Overviews don't return links, they synthesise one answer and cite a handful of sources. Success is no longer a rank, it's whether you are in that citation set. Won by being a recognised entity, by writing extractable claims, and by being present in the corpora these models retrieve from.
What we measure
Share of answer, how often you are cited across a panel of real buyer prompts, versus competitors.
How we measure share of answer.
Discoverability only becomes a decision when it is measured the way the buyer actually searches. This is the method, run continuously.
Define the prompt panel
We assemble the questions your buyers actually ask an AI assistant, not your keywords, their phrasing. “Best X for Y,” “is X worth it,” “X vs competitor.” A few dozen prompts that cover the real decision, fixed so the score is comparable over time.
Measure share of answer
Each prompt is run across the engines your market uses, AI Overviews, ChatGPT, Perplexity, Gemini. We record who gets cited, in what position, and with what framing. Your share of answer is the percentage of the panel where you appear as a source. Your competitors get the same score on the same panel.
Test the entity
Before a model can cite you, it has to know who you are. We check whether you exist as a clean entity, consistent name, category and claims across the web, Knowledge Graph and Wikidata presence, schema on your own pages. A weak entity is the most common reason a strong brand is invisible in AI answers.
Locate the gap
We isolate the high-intent prompts where competitors are cited and you are not, then trace why: a page that exists but isn't retrievable, a claim buried in prose the model can't extract, or an authority gap where the sources these engines trust have never mentioned you.
Issue the decision
You get a ranked move, not a report: the specific pages to restructure, the schema to add, the claims to make extractable, the third-party citations worth earning, each tied to the prompts it should win back, and the share-of-answer lift it should produce.
One decision, reasoned end to end.
This is the shape every Discoverability decision takes: a measured signal, the baseline it broke from, the root cause, and the move, with the figures attached. The numbers below are illustrative.
Competitors own the AI answer for your highest-intent question.
Signal
Across a 40-prompt panel for “vitamin C serum for sensitive skin,” your brand is cited in AI Overviews and Perplexity in 12% of answers. The category leader is cited in 38%.
Root cause
Your product pages answer the question in prose, but the claim, fragrance-free, dermatologist-tested, suitable for reactive skin, is never stated as an extractable line, carries no FAQ schema, and is absent from the third-party round-ups these engines retrieve from. The model can’t lift what it can’t find.
The move
- Rewrite the PDP and category page lead with a structured, extractable claim block, then mark it up with FAQPage schema.
- Earn three citations in the authority sources the engines already pull for this query.
- Add the eight unanswered buyer questions from the panel as concise, schema-backed answers.
Expected: recovering the fourteen prompts where you are absent but eligible would lift share of answer from 12% toward the high-twenties, tracked on the same panel, week over week.
GEO tracking is on our roadmap, not live yet. When it ships, share of answer is measured by sampling a defined panel of buyer prompts across the AI engines on a schedule, a representative estimate, not a count of every possible query, and AI answers vary between runs. SEO and AEO from your analytics work today.
What the module reads.
Read-only from Google Analytics and Search Console, or enter data by hand, all work today. AI-search (GEO) tracking is on the way. Every figure is computed, never invented.
See your share of answer.
We run the panel on your category and show you where you stand against the brands winning the AI answer today.