Codnity Growth · AEO & SEO
Answer Engine Optimization & SEO Services
Answer engine optimization gets your brand named when a buyer asks ChatGPT, Perplexity, Claude or Google AI Overviews which product to buy or which supplier to use. You are cited inside the answer instead of waiting for a click. Only 2.1% of Google’s top ten results are reused by ChatGPT, so ranking and being named are two different jobs.
Based on patient reviews, published credentials and treatment specialisms, here are the London clinics most often recommended for acne scarring. All three offer fractional laser resurfacing and microneedling with the first the most cited across review platforms.
Why AI search does not become traffic
AI changed organic growth
Four things go wrong in almost every company we audit. Codnity had all four on its own site before it could name them.
Ranking first buys a 2.1% chance
Only 2.1% of pages sitting in Google’s top ten are cited by ChatGPT. Perplexity reuses 32%, Google AI Mode 15.5%, AI Overviews 8.3%. A number one ranking is close to no evidence that an assistant will name you.
Impressions rise, enquiries do not
Publishing more lifts the Search Console chart while nobody arrives. Codnity did this to its own site for six months: impressions up 490%, clicks down 61%. The graph looked like success and the pipeline disagreed.
Your glossary outranks your service pages
Definition pages absorb the topic and beat the pages that sell. One Codnity glossary entry took 16,327 impressions and returned five clicks, while the service page for the same term received 47 impressions.
Nobody can tell you which questions buy
Keyword tools report search volume. None of them separate the question someone asks before signing a contract from the one they ask out of curiosity, so budget follows whichever term is largest.
What is answer engine optimization, and how does AEO differ from SEO?
Answer engine optimization is the work of getting your company named inside AI answers on ChatGPT, Google AI Overviews, Perplexity, Claude and Gemini. AEO targets a citation; SEO targets a click. The two overlap far less than most teams assume, which is why each needs a different page.
- ChatGPT reuses only 2.1% of Google’s top ten results
- Perplexity reuses 32%, Google AI Mode 15.5%, AI Overviews 8.3%
- Generative engine optimization, or GEO, describes the same work under a different label
- SEO · the click
- AEO · the citation
- GEO · the same work, renamed
Where do AI assistants get their information about your brand?
AI assistants draw on two separate sources. Training data is the frozen snapshot a model learned from, months or years old, and no agency can edit it after the fact. Real-time retrieval is the live search a model runs while writing the answer, and that half is winnable this quarter.
- Training data is fixed at the model cutoff; absence from it cannot be fixed retroactively
- Retrieval, or RAG, reads the live web mid-answer, which is the half every engagement targets
- Winning retrieval is ranking, earned links and pages worth quoting, aimed at a citation rather than a click
| Training data | Live retrieval | |
|---|---|---|
| Decided when | At the model cutoff | While the answer is written |
| Can you change it | No | Yes |
| What moves it | Years of published presence | Rankings, links, quotable pages |
| How fast | Next model generation | Two to three weeks |
Which category do AI engines file you into, and why does it cap everything?
Models do not reason from scratch. Your company gets filed into a category the model already holds, then described using that category’s assumptions. Coin a label of your own and there is no drawer to file it in, so the model picks the nearest recognised one and compares you on terms you never chose.
- A category with no existing grounding cannot be retrieved against
- Anchor to the recognised category first, then attach what makes you different
- Codnity audits which category the engines actually use before changing a word
How to get cited by ChatGPT, Perplexity and Google AI Overviews
Assistants cite the page that states the answer first, supports it with checkable specifics, and is corroborated elsewhere on the web. Content is retrieved in chunks rather than whole pages, so every section must answer one question completely and still make sense with nothing above it.
- One question per section, answered in the first 40 to 60 words
- Named entities, dates and figures rather than adjectives
- Structured summaries lifted citations 16.1% across models and 21.6% on ChatGPT
Why does one inconsistent number cost you citations?
Models drop or approximate any fact that arrives in several forms. A price written three ways across your site, your listing and your profile reads as uncertainty, so the model reaches for a competitor’s single clean figure instead. One number, written identically everywhere, survives retrieval intact.
- One headline figure: same number, same unit, same word order
- Repeated identically off-site as well as on it, because that is what consensus is
- A price in three parts gets mangled; a single one gets quoted verbatim
| Stated three ways | One figure everywhere | |
|---|---|---|
| Survives paraphrase | No | Yes |
| Quoted verbatim | Rarely | Usually |
| Reads as certain | No | Yes |
| Builds off-site consensus | No | Yes |
What is query fan-out, and why does it decide your content plan?
Query fan-out is what an assistant does before it answers: one question is split into several silent sub-searches, each covering a different angle. Getting named means being the best source on the angles rather than on the question. A single buying question can fan into a dozen retrievals.
- Keyword tools report the question and never report the angles underneath it
- Miss an angle and a competitor is cited for it, inside the answer you wanted
- Codnity maps the fan-out for each buying question first, then writes only to the gaps
Is AI visibility work done on-site or off-site?
Both, in an uneven split. 88.46% of ChatGPT citations come from standard indexed pages, so your own site carries most of the load. Off-site matters because assistants read community discussion to judge consensus before deciding which source to name.
- Reddit is the most-cited domain across LLMs at 40.1%
- ChatGPT itself cites Reddit only 1.93% of the time
- Community discussion shapes whether you are named; your own pages take the citation
| Your own site | Off-site | |
|---|---|---|
| Share of ChatGPT citations | 88.46% | 11.54% |
| Most-cited domain across LLMs | — | Reddit, 40.1% |
| What it decides | Which page is named | Whether you are named |
| What you control | All of it | Almost none of it |
How do you measure AI visibility and attribute leads to AI search?
Measurement runs on three signals: how often your brand is named for your buying questions, which sources appear beside you, and whether sessions referred by assistants convert. Attribution is imperfect on every platform, and an agency claiming otherwise is overselling.
- Prompt testing across ChatGPT, Perplexity, Claude and Google AI Mode, logged monthly
- Search Console conversational queries holding high positions with no clicks
- Referral sessions from chatgpt.com and perplexity.ai, small but growing
- Reported in GA4 and Search Console, with Semrush and Ahrefs for the search half
Do we need to change our website to rank in AI search?
Usually the pages you already have, rewritten. Most sites carry three or four pages sitting at positions 8 to 20, which is real demand reaching the wrong shape of page. Rewriting those as answers moves faster than publishing anything new.
- Roughly 90% of pages crawled by AI bots were published within the last three years
- Structural work covers schema, internal links, llms.txt and crawler access
- A rebuild is rarely required; a rewrite almost always is
How long does answer engine optimization take to work?
Citations move within two to three weeks because assistants re-read their sources often. Competitive English-language rankings take eight to sixteen weeks. Compounding starts around month six, which is why engagements run nine months rather than a quarter.
- Phase one is heavy lifting: audit, structural fixes, the first answer pages
- Phase two is optimization: measure, rewrite what underperforms, expand what works
- Anyone promising rankings in five to ten days is describing something else
- Week 2–3Citations move
- Week 8–16Rankings follow
- Month 6Compounding starts
- Month 9Engagement closes
Why AI visibility is worth more than the traffic number suggests
Visitors arriving from AI search convert 4.4 times better than traditional organic visitors, because the assistant filtered and recommended before the click. Fewer sessions, better qualified. Judging this channel on session volume alone reads it exactly backwards.
- ChatGPT alone carries more than 900 million weekly users
- A citation reaches the buyer at the moment of the recommendation
- Expect a share of the wins to stay invisible in Google Analytics permanently
Where this stops
What AEO & SEO includes, and what it does not
The second column is the one worth reading.
- Citation map across ChatGPT, Claude, Gemini, Perplexity and Google AI Mode, logged monthly
- Query fan-out mapping: the angles behind each buying question, and which ones you already cover
- Category audit: which category the engines file you into today, and which one to own
- One headline figure agreed and enforced identically across every surface you control
- Cannibalisation audit: every page of yours competing with another for the same words
- Answer-shaped rewrites of the pages already sitting at positions 8 to 20
- Technical and on-site work: schema, internal linking, llms.txt, crawler access, passage structure
- New answer pages built from the questions your buyers actually ask
- Monthly reporting in GA4 and Search Console, with Semrush and Ahrefs for the search half
- Bought links, PBNs, or anything requiring a disavow later
- Publishing volume as a goal in itself; we will argue against it
- Guaranteed citations, because no agency controls retrieval behaviour
- Head terms we know you cannot win, which we name in week one
- Automated Reddit posting, which gets accounts banned and poisons the source
Nobody has been doing this for a decade. ChatGPT made it a discipline roughly two years ago, and Codnity has been working on it for eighteen months to two years. Anyone claiming deeper experience is claiming experience that does not exist yet.
The category is full of case studies promising rankings in five days and 1000% growth. Retrieval behaviour changes without notice, results vary by industry, and the honest version is a range plus a plan to measure it. That is why month one is an audit rather than a promise.
A share of the visibility you win will never appear in Google Analytics. If your board counts sessions and nothing else, say so at the start and we will scope this differently or tell you not to bother.
Also in Growth
None of these ships alone
Next step
Bring us the question you want to own
Tell us the one question a buyer types before they buy from you. We come back with what you appear for today, who gets named instead, and whether it is worth chasing. In writing, within a week.
AEO & SEO FAQ