Refresh or Rebuild? The AI Website Diagnostic

Why Websites Built Before 2026 Underperform

The share of consumers using AI tools to find local businesses rose from 6% in 2025 to 45% in 2026 — a 7.5x increase in twelve months. Google AI Overviews now appear on up to 68% of local search queries, and when one appears the average zero-click rate is 83%. Meanwhile only about 1.2% of local businesses are ever recommended by an AI engine.

The practical consequence for a medical spa: a website can rank well, look current, and still be structurally unreadable to the systems that answer "best med spa near me." That is a legibility problem, not a design problem, and the two require different fixes.

Over 75% of med spa website traffic is mobile. Clinics that rebuild mobile-first commonly report large conversion swings — one documented case moved mobile bounce from 72% to 38% and increased mobile bookings 140%. Booking paths longer than three taps measurably reduce completed appointments.

SEO vs AEO vs GEO — What the Terms Actually Mean

These three are used interchangeably by people selling them, but they optimize for different outcomes and the difference determines where budget should go.

SEO (Search Engine Optimization)

Goal: earn a position in a list of links. Still sends traffic, but it competes for the clicks that remain after an AI Overview has already answered the question.

AEO (Answer Engine Optimization)

Goal: be the passage that gets extracted. On-page work — question-shaped headings, the direct answer within the first 100 words, clean semantic HTML, and content a machine can lift without inference.

GEO (Generative Engine Optimization)

Goal: be the source a generative model cites when composing an answer. Off-page and structural — entity clarity, schema, and corroboration across the web. It cannot be achieved through writing alone; it must be built and confirmed elsewhere.

What Actually Gets a Website Cited by AI

AI does not rank a business. It decides whether the business is worth using as a source. Being cited as a primary source across ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews requires two layers: usability and foundation.

Usability — can an engine extract and recombine the content? This requires a direct answer in the first 100 words, question-shaped H2 headings that match how patients actually phrase things, topic clusters rather than isolated pages, and server-rendered HTML, because most crawlers do not execute JavaScript. Content that only appears after a JavaScript framework mounts does not exist to them.

Foundation — every model asks two questions. Trust: does the web corroborate this business? That requires external citations from authoritative sources, real reviews using consistent language, and the same name and story everywhere the business appears. Entity: does the engine understand who this business is? That requires Organization, Person, Service, and LocalBusiness schema, sameAs links tying verified profiles to a single entity, and named authors and providers with real credentials.

The common expectation is that better content is the answer. What actually matters is alignment across everything.

Why AI Engines Go Silent About a Business

When an AI engine does not recommend a business, the common assumption is that the business is ranked low. More often the model could not confidently resolve who the business is, and an uncertain model does not guess — it stays silent and names a business it is sure about. Inconsistent or unverifiable brand data pushes models toward silence rather than citation.

Three technical conditions cause this, and none of them are visible from the front of a website. Inconsistent NAP data — where the Google Business Profile, the website, and third-party directories disagree on name format, suite number, or phone — creates entity ambiguity that prevents an engine from resolving multiple records into one canonical business. Broken links and dead pages consume crawl budget that would otherwise be spent reading the pages the business wants cited, and a site full of 404s reads as unmaintained, which is itself a trust signal. Navigation an engine cannot follow leaves valuable pages effectively orphaned, because crawlers discover content by following links.

This is why a site with substantial technical debt is a remediation project rather than a content refresh. The work of reconciling entity data across the web, repairing link structure, and rebuilding navigable architecture is separate from, and usually larger than, the work of writing better pages.

E-E-A-T for Medical Aesthetics Websites

Google formalized Experience, Expertise, Authoritativeness, and Trust for its human quality raters. Answer engines operationalized the same criteria and check them mechanically, at scale, on every query. Medical aesthetics is a your-money-or-your-life category, where that bar is set highest.

Experience

Has this business actually done the thing? Named injectors with real case volume, before-and-after documentation with treatment details rather than stock photography, and a point of view that could only come from doing the work. An engine quotes a real person before it quotes a logo.

Expertise

Is the depth demonstrated or merely implied? Credentials stated explicitly — PA-C, NP, MD, board certification, device training — and content that answers the hard questions: contraindications, who is not a candidate, and what goes wrong. Easy answers are everywhere; hard ones get cited.

Authoritativeness

Does anyone else confirm it? The same name, credentials, and story across Google Business Profile, RealSelf, Allergan, and state licensing boards, tied together with sameAs links in schema. AI trusts what more than one independent source confirms.

Trust

Is any of it verifiable? Real pricing or honest ranges, reviews that repeat consistent language about results, transparent policies, and an address that matches everywhere. Vague claims read as risk; specific claims read as accountable.

Six Things That Won a Live AI Recommendation Query

StrataVera ran a live recommendation query in the medical aesthetics category and analyzed why the model selected the practice it did. Six factors carried the result, and none of them was the website's visual design.

A specific, verifiable claim

The winning practice led with "top 250 of 33,000+ providers." A number a model can repeat outperforms an adjective it has to interpret.

The injector named, with credentials

PA-C, NP, or MD stated directly on the page. AI quotes a real person before it quotes a brand.

A stated point of view

A treatment philosophy such as "less is more" gives the model a quotable line. A service list does not.

Reviews repeating the same language

Consistency reads as proof. Scattered praise reads as noise, even in higher volume.

One name and story across the web

Google, RealSelf, and Allergan all in agreement. AI trusts what multiple sources confirm.

A site the model could actually read

Real text rather than images. Stated specialization beat visual polish — many beautiful sites cannot be read at all.

The Three Verdicts

Refresh

The foundation is sound. The site needs a modernized surface, a shorter booking path, and the answer layer that machines read — not a teardown.

Rebuild

Improvements no longer compound because the underlying structure was designed for a different web. A new build on a modern stack costs less over eighteen months than continuing to pay the maintenance tax.

Replatform

The platform is the larger problem — domain, hosting, email, booking, and analytics scattered across vendors with no clear owner. Requires a new foundation the business actually controls.

What the Diagnostic Measures

Foundation risk

Build age, mobile booking friction, content editability, mobile load speed on cellular data, and whether website-sourced inquiries are measurable at all.

AI-invisibility risk

Whether AI engines currently name the business, whether the site answers real patient questions in plain text, whether service content exists as text rather than images, and whether structured data identifies the business to machines.

The AI-Ready Website

StrataVera builds websites designed to be read twice: once by the patient deciding whether to trust the practice, and once by the machine deciding whether to recommend it. Managed hosting keeps the site in that state.

Answer-engine architecture

Structured data, entity markup, and a machine-readable content layer so ChatGPT, Perplexity, Claude, and Google AI Overviews can read and cite the practice accurately.

Three-tap mobile booking

Wired directly into Boulevard, Vagaro, Zenoti, or GoHighLevel with no dead ends and no call-during-business-hours requirement.

A real answers layer

Pricing, downtime, and candidacy questions answered in plain text — the content shape that earns AI citations.

Owner-editable content

Practice owners change their own prices and offers in minutes without a support ticket, or the managed hosting plan handles it.

Performance budgets

Speed targets set at build time and monitored after, because load time on cellular data is where conversion is won or lost.

Closed-loop attribution

Every inquiry tagged to its source so marketing spend follows evidence rather than instinct.

Managed AI-Ready Hosting

Optional ongoing hosting, SSL, backups, uptime, performance budget, quarterly content refresh, and monthly citation tracking showing which engines name the practice.

The Citation Engine — Ongoing GEO Content

Not every practice needs a new website. When a site is already well built, the constraint is usually that it has published nothing worth quoting. An answer engine can read the site perfectly and still have no reason to name the business, because nothing on it answers the question being asked. Structure without substance does not earn citations.

The Citation Engine is StrataVera's ongoing GEO content program. It builds and continuously feeds the answer layer a site is missing, on the principle that citation is a standing position rather than a project that finishes. It is the recommended path for practices whose diagnostic returns Refresh, and for any practice that has recently rebuilt.

Topic clusters rather than isolated pages

One page about a treatment is a mention. A dozen interlinked pages covering candidacy, cost, downtime, aftercare, and comparisons is topical authority — and authority is what gets quoted.

Question-shaped articles

Built around the phrasing patients actually use, with the direct answer placed in the first hundred words where a model will extract it.

A structured Q&A and glossary layer

The honest questions — pricing, contraindications, who is not a candidate — published as structured Q&A with FAQPage schema and defined terms marked up for machine parsing.

Published under the practice’s own clinician

StrataVera drafts and structures the content; the practice’s medical director or lead injector reviews and approves each batch, and articles publish with that provider’s byline and credentials attached, marked up as author and reviewedBy in schema. This is both the compliance step for a your-money-or-your-life category and the strongest available citation signal, because answer engines weight named, credentialed authors over unattributed brand content. Nothing publishes without provider approval.

Produced on a dedicated content platform

StrataVera runs its own content generation and publishing system, which supports a publishing cadence most individual practices cannot staff for and requires no writing from the client.

Who Builds These

StrataVera is led by founder Christy Rexroth, whose background is operational rather than agency. She ran a Diamond-level Allergan medical spa — the top tier of the Allergan partner program — and spent more than a year working directly with senior Allergan marketing leadership on how aesthetics practices actually grow. Before founding StrataVera she came up through clinical aesthetics with iS Clinical, working on practice floors with hundreds of owners.

That operator background matters for AI visibility specifically. The content that earns citations in medical aesthetics is the content that answers what a patient is too embarrassed to ask out loud — real pricing, downtime, whether she is even a candidate, what happens when it goes wrong. That content cannot be produced from a keyword tool; it comes from having been in the room.

Diamond-level Allergan practice

Operated a med spa at the top tier of the Allergan partner program.

Allergan marketing leadership

More than a year working directly with senior Allergan marketing leaders on aesthetics practice growth.

iS Clinical, clinical aesthetics

Years on practice floors with hundreds of medical spa and practice owners.

The AI Visibility Session

The entry point to working with StrataVera is a complimentary 30-minute AI Visibility Session rather than a proposal. Before the call, StrataVera runs the practice’s real market queries across ChatGPT, Perplexity, Gemini, and Google AI Overviews — the questions patients actually ask, such as "best med spa in [city]" or "where to get Botox near me."

On the call, the owner sees the actual answers those engines return, including which competitor is being named instead of them and the specific factors on that competitor’s site earning the recommendation. The owner leaves with a written, prioritized sequence — what to fix first, what can wait, and what to leave alone — which is theirs to keep whether they engage StrataVera, engage someone else, or hand it to whoever already maintains the site.

Common Questions

Is this just SEO with a new name?

No. Traditional SEO optimizes for ranking in a list of links. Answer-engine optimization targets being cited inside a generated answer — a different structure, content shape, and measurement. StrataVera does both, because search still sends traffic, but only one channel is growing.

What if only a small fix is needed?

Then the diagnostic returns Refresh and the recommendation is to spend less. Roughly a third of scored sites come back Refresh.

What is the AI Visibility Session?

A complimentary 30-minute session. Before the call StrataVera runs the practice’s real market queries across ChatGPT, Perplexity, Gemini, and Google AI Overviews. On the call the owner sees which competitor is being named instead of them and why. They leave with a written prioritized sequence regardless of whether they engage StrataVera.

How is pricing determined?

Scope follows the diagnosis rather than a package menu, and StrataVera does not quote before assessing the site — scoping a Refresh as a Rebuild wastes the client’s money. A written number and matching scope follow the AI Visibility Session.

Is moving hosting required?

No. Hosting is a separate decision. The build, domain, accounts, and code belong to the client either way.

Who is this for?

Medical spa owners, multi-location clinic operators, and specialized service practice founders — particularly those whose website was last rebuilt more than two years ago.