You are probably reading this with several other tabs open, each one an agency promising to get you cited by ChatGPT. This page is going to be different from those tabs in one specific way: before we tell you what we sell, we are going to tell you what the evidence actually supports - including the parts that make our own category look smaller and harder than the other tabs claim.

What answer engine optimization is.

Answer engine optimization is the practice of making your identity, your evidence, and your content legible to the systems that now answer questions directly - AI assistants like ChatGPT and Claude, AI-generated summaries in Google and Bing, voice interfaces. When a prospective client asks about your field, the aim is that the answer includes you, and gets you right.

The reason it exists as a discipline: the click economy is thinning. Pew Research Center found users click a traditional search result about 8% of the time when an AI summary is present, versus 15% without one. And presence in those answers is not the same game as ranking - Ahrefs found only about 12% of AI-cited pages also rank in Google's top ten for the query. A practice can be invisible in rankings and still become the cited answer. We have watched it happen, and documented it, which is what the case study below is for.

If you want the full definitional treatment - AEO versus GEO versus SEO, and which of the circulating statistics survive checking - we wrote it up separately: AEO vs GEO vs SEO: What the Acronyms Mean, and What the Evidence Says.

How answer engines actually choose.

Understanding the mechanism protects you from most of the bad pitches, so here it is, in the version the evidence supports, and where it runs out.

When an AI assistant composes an answer, it retrieves candidate material, weighs it, and synthesizes. What the evidence says about that weighing: citation correlates with semantic relevance to the specific question asked (how directly and completely a page answers the actual prompt) more than with brand size or ranking history. Depth on the narrow question beats breadth on the general topic. This is why a one-person practice with the most specific, most verifiable answer in a narrow field can be cited while larger competitors with stronger domains are not.

Two more mechanism facts worth carrying into any sales conversation. The engines may differ from each other: across ChatGPT, Perplexity, Copilot and Gemini, cited pages were about 26% newer than Google's organic results in a 2025 Ahrefs study, while Google's AI Overviews cited older pages in one 2025 study and newer pages in another. Ask which engine a freshness tactic was tested on. And being retrieved is not being cited: Ahrefs found ChatGPT cited 1.93% of the Reddit pages it retrieved, and of the pages ChatGPT retrieved but did not cite, 67.8% were Reddit's. Ahrefs cautions that some of those pages may never have been opened. An honest provider will tell you which of the two outcomes a given piece of work is likely to produce.

The mechanism is verifiable identity plus specific, evidence-backed answers, weighted by systems that change quarterly. That is what you are actually buying when you buy AEO. Everything else on this page follows from it.

What the industry claims, and what survived verification.

We adversarially checked the most-circulated claims in this space against their primary sources and published the results as a standing, dated register. Several widely repeated claims do not hold up as sold. A sample of what you will encounter in those other tabs:

The pitch you'll hearWhat the evidence says
"Schema markup gets you cited by AI."Correlated, not causal. A matched study of 1,885 pages found no citation lift from adding it - AI Overview citations even declined slightly. Schema matters for a different reason: it makes claims machine-checkable.
"GEO tactics boost AI visibility up to 40%."The figure is a lab benchmark result from the 2024 GEO paper. No study has shown it in the field. Do not build a budget on it.
"We'll make you visible everywhere AI looks."Only ~32% of the pages ChatGPT cites most are realistically influenceable by content or brand work at all. The rest goes to Wikipedia, homepages, and reference sites no agency touches. That was a September 2025 snapshot of the most-cited pages overall, not a forecast for your field. We scope against your specific questions.
"AI engines reward consensus - we'll get you mentioned everywhere."No one has measured the "consensus layer" theory. What citation is measured to correlate with is semantic relevance to the specific question asked.

The full list, with sources, lives in the AI-Search Mythbusting Register and gets updated as new claims get checked. We maintain it in public because "we're the ones who checked" is the entire basis on which we'd ask you to trust the rest of this page.

What we actually build.

An engagement is a rebuild: a new website for your firm with the identity, evidence links and editorial structure built in, over six to eight weeks. If you want to keep your current site, we are probably not the right firm.

One finding cuts against part of this work, and we report it. Neither our own law-firm study nor Ahrefs' 1,885-page study found that structured data on its own earns citations, and Google says it is not required for its AI features. We build it because it makes a firm's claims checkable by machine. Two items on this list have some measured support: answering the specific questions clients ask, and being retrievable by the engines' own searches (one preprint, ChatGPT only). The entity graph, evidence linkage and discovery files are our bet that machine-checkable identity helps. Nobody has measured it. We build them because they make your claims checkable, and the cost is small next to the rebuild.

Clarion does not sell AEO as a bolt-on, a schema plugin here and a content tweak there. The work is infrastructure, built once and built to be verified. The complete technical walkthrough is public in How AI Verifies You're Real; the short version of what an engagement produces:

  • A diagnostic of how AI engines describe you today. Dated captures of what ChatGPT, Perplexity, and Google's AI surfaces actually say when asked about you and your field - including where they get you wrong, which is the failure mode few providers measure.
  • A canonical entity graph. Organization, principal, credentials, and services declared once, consistently, with stable identifiers - the machine-readable spine everything else references.
  • Evidence linkage. Every significant claim wired to something outside the page a machine can check: regulatory records, registries, credential bodies, dated documents. This is the layer the schema-plugin crowd skips, and it is the one that separates syntactic validity from verifiable authority.
  • Discovery infrastructure. Sitemap, AI-aware robots.txt, llms.txt, and full-content files - the surfaces AI crawlers actually read, kept current as the site grows.
  • Answer-shaped editorial architecture. The questions your field is actually asked, answered at depth under your name, structured so both a reader and a machine can extract the answer.
  • Dated before/after measurement. The same citation captures, re-run on a recurring cadence, so the engagement's effect is documented rather than asserted - and so drift (an engine describing you wrongly) gets caught.

The dimension few providers measure: fidelity.

Most AEO providers measure presence: were you mentioned, were you cited, how often. Presence is half the story. AI engines sometimes blend a source's real content with other material into a composite that is quietly wrong, then present it as the original. A page can be cited and misrepresented in the same answer.

For a professional practice that matters more than a missed mention. An engine that gets your pricing, your credentials, or your scope of service wrong is misinforming the exact person who asked about you, at the exact moment they were deciding whether to call. So Clarion's recurring captures measure two things: whether the engines mention you, and whether what they say matches what your own content says. When it doesn't, the drift itself tells us which claims need stronger external verification. That second measurement is the part you will have trouble buying elsewhere, and it is the part a referral-driven practice actually needs.

Who this is for. And who it isn't.

This work fits firms whose people are paid for judgment in narrow, contested fields - law firms from one practice group to many offices, registered investment advisers, other professional-services firms where the firm owns the reference and its people carry the bylines. It fits where being the cited authority is worth real money: high client lifetime value, a niche someone can actually own, decisions clients research before making. It does not fit a business competing on price, a firm wanting schema bolted onto an existing Wix or Squarespace template, or anyone whose clients don't research before buying. We say this before the diagnostic conversation so neither of us spends an hour discovering it there. The fuller fit criteria are on the approach page.

What we won't sell you.

No guaranteed citations or placements - AI engines' behavior is not controllable from the outside, and anyone guaranteeing it is guaranteeing something they don't control. No schema-as-magic. No "be everywhere AI looks" - in September 2025 only about a third of the most-cited pages were the kind content can influence, and we scope against your specific questions. No monthly content mill - we build editorial systems, not word counts. And no promises built on statistics that failed verification, even when they'd make this page more persuasive. If the honest ceiling is lower than the other tabs told you, we'd rather you hear it here, before you've paid anyone.

The proof we can show you.

Our primary demonstration is deliberately small and checkable, and it is the founder's own business: Clarion's seven-layer method applied to a small Lagotto Romagnolo breeding program with no marketing budget - followed by eleven live AI-engine transcripts citing it by name, including an unprompted "this would probably be my first call." Every capture is dated, and the whole thing is written up as The Smallest Possible Proof. We also publish a 130-firm audit of one profession's AI-search infrastructure (The AI Visibility Gap) and score our own homepage with the same Schema Inspector we point at everyone else's.

What we do not have: a wall of logos. The studio is new, deliberately small, and the reference work is disclosed for exactly what it is on the work page, including how each engagement came to us. You should read that page with the same skepticism this page has been teaching you to apply.

How an engagement runs.

Four phases over six to eight weeks, each with a defined output and a pause to confirm we are building what you actually need.

Foundation, weeks one and two. The diagnostic captures of how AI engines describe you today, a positioning audit, and the entity-graph architecture designed for your practice specifically. This phase ends with you seeing, in writing, exactly what will be built and why.

Build, weeks three through five. The site, the structured data, the evidence linkage, the editorial system, and the discovery infrastructure. Clean-slate and hand-built; nothing retrofitted onto a platform. The complete seven-layer specification this follows is public in How AI Verifies You're Real.

Launch, weeks six and seven. DNS migration, schema verification, and discoverability confirmation across both traditional and AI search. Quiet and deliberate; no broken redirects.

Handover, week eight. Documentation, editing conventions so the system can grow without breaking, the baseline citation captures filed, and the recurring fidelity checks, which run through the first year. After that, measurement continues on an optional retainer. The full methodology, including who we are and aren't a fit for, is on the approach page.

Investment.

Engagements begin at $35,000; most land between $35,000 and $65,000, scoped to the depth of the editorial system, the number of principal entities, and the complexity of verification linkage. You will notice the other tabs don't publish numbers. We do, for the same reason the register exists: you should be able to check us against what we've said before you ever talk to us. A diagnostic conversation establishes scope before either of us commits to anything.

Questions worth asking any AEO provider.

What is answer engine optimization?

The practice of making your identity, evidence, and content legible to AI systems that answer questions directly, so you appear - accurately - in the answers your prospective clients receive. Presence in the answer, not position in a list, is the unit of victory.

How is it different from SEO?

Different game, different rules. In one 2025 study, about 12% of AI-cited pages also ranked in Google's top ten, averaged across five engines. Studies that also count the engines' follow-up searches find more overlap. Ranking helps, and citation can be won without it.

Does schema markup get me cited?

Not by itself. A matched study of 1,885 pages that were already heavily cited found no lift from adding it, and it cannot say what schema does for pages the engines have not found. Google says structured data is not required for its AI features. We build it because it makes your claims machine-checkable. Ask any provider who leads with schema what else they build.

How much of this is actually winnable?

About a third of the pages ChatGPT cited most in a September 2025 study were realistically influenceable. The mix has shifted since, so we scope against your specific questions, not that number.

Can you guarantee I'll be cited?

No, and no one honestly can. We contract for infrastructure and dated measurement, not placements. Ask every provider this question and watch what they do with it.

What's included?

Diagnostic captures, canonical entity graph, evidence linkage, discovery infrastructure, answer-shaped editorial architecture, and recurring before/after measurement through the first year, described under What we actually build on this page.

What does it cost?

From $35,000; most engagements land between $35,000 and $65,000, scoped in a diagnostic conversation first.

How fast will I see results?

AI engines re-crawl and re-synthesize on their own schedules; we assess movement over 90-day windows with dated captures rather than promising timelines. The measurement is part of the deliverable.

My clients come by referral. Do I need this?

Some referred prospects ask an AI assistant about you before they call. We have no measure of how many. If an assistant describes you thinly or wrongly to a referred prospect, the referral leaks. This work makes the machine's answer match the reputation the referral was built on.