AI search visibility and GEO assessment
Find out how AI answers represent your company.
I review your priority pages and a defined set of customer questions, then separate observed issues from hypotheses and give you a practical improvement plan.
The facts
- Fixed fee
- base or extended tier
- 3 weeks
- kickoff to handoff
- 25 pages
- priority pages, one domain
- 60 observations
- 15 questions × 2 platforms × 2 runs
Base tier: 25 pages, 15 questions, two named answer platforms, one market and language. Extended tier: up to 75 pages, 50 questions, three platforms, and two competitor benchmarks. Monitoring and implementation are scoped separately.
Use this when
- Someone has asked how your brand shows up in AI answers and nobody can answer with evidence.
- A vendor is offering a proprietary visibility score you can't defend in a meeting.
- You suspect AI answers describe your offer, category, or pricing inaccurately.
- You need to know which pages and claims to fix first, and in what order.
You leave with
A dated baseline of how named answer platforms respond to priority customer questions, page-level evidence, findings labeled observed, likely, or unknown, a prioritized improvement plan, and the reusable question set your team needs to rerun the baseline later.
This is useful even if you don't yet see measurable AI referral traffic. The assessment starts with how AI answers represent the brand, not with a minimum traffic threshold. Sometimes called GEO, or generative engine optimization.
Observation to finding
What an AI answer said, what your page says, and what to change.
Observation · dated
- Question
- “Does [Company] integrate with HubSpot?”
- Platform
- Platform A, fresh session, run 1 of 2
- Date
- 2026-09-02, 14:10 ET
“[Company] does not appear to offer a HubSpot integration.” No citation to the company’s site.
Source page · reviewed
- Page
- /integrations/
- Access
- Rendered check: integration list loads only after a client-side script runs
- Entity
- “HubSpot” appears once, in an image alt
Excerpt of the page as fetched without scripts, showing the integration list absent from the delivered HTML.
Finding · and the change
ConfirmedPriority: highThe integration list isn’t in the HTML a non-rendering fetch receives, and the only text mention of HubSpot is an alt attribute. Both runs on both platforms returned “no integration” or “unclear.”
Change: render the integration list in the delivered HTML with each partner named in text. Add a short plain sentence per integration. Recheck in four weeks with the same 15 questions.
Exactly what's covered
Scope you can count.
- 25
priority pages
On one domain, chosen with you at kickoff by business importance and opportunity, not only current traffic.
- 15
customer questions
Representative of what buyers ask. One market and one language.
- 2
named answer platforms
Agreed in the signed scope. Each question runs twice per platform in fresh sessions.
- 60
dated observations
Each recorded with question, platform, date, answer, and any sources the answer cited.
What I review
Four checks on every priority page.
Google documents ordinary SEO foundations for its AI features, not special AI markup. Other platforms don't all read pages the same way. The review checks what can be checked.
| Check | The question it answers |
|---|---|
| Access | Can the page be fetched and read without running scripts? What does a rendered fetch see that a plain one doesn't? Any blocks in robots or headers? |
| Content | Does the page answer the customer question in plain text, in the words a buyer would use? Or is the answer implied, in an image, or on a page nobody links to? |
| Entity | Is it clear who the company is, what it does, and how it's named across the site and elsewhere? Consistent names, consistent claims. |
| Evidence | Are the claims on the page supported somewhere an answer engine can find? Original data, named sources, clear authorship. |
Access
The question it answersCan the page be fetched and read without running scripts? What does a rendered fetch see that a plain one doesn't? Any blocks in robots or headers?Content
The question it answersDoes the page answer the customer question in plain text, in the words a buyer would use? Or is the answer implied, in an image, or on a page nobody links to?Entity
The question it answersIs it clear who the company is, what it does, and how it's named across the site and elsewhere? Consistent names, consistent claims.Evidence
The question it answersAre the claims on the page supported somewhere an answer engine can find? Original data, named sources, clear authorship.
How findings are labeled
Observed, inferred, or undetermined. Never blurred.
The most useful thing in the report is knowing which findings you can act on today and which ones are a good guess.
Seen directly in the observation and the page. Repeated across runs. Act on it.
Consistent with the evidence but not directly observable. The report says what would confirm it.
Answers varied between runs, or the cause can't be isolated. Logged, not guessed at.
Outputs
Four things you keep.
Observation log
All 60 dated observations with question, platform, answer, and cited sources.
Page review
Access, content, entity, and evidence checks for each of the 25 pages.
Prioritized changes
Each finding labeled, tied to its evidence, with the specific change and who'd make it.
Working handoff
A session with whoever owns the site, plus the question set so you can rerun the check yourself.
Limits
What this can and can't tell you.
A baseline, dated. Useful because it's specific, not because it's permanent.
- Answers vary
- Repeated prompts can produce different answers. Two runs per platform show consistency; they don't prove what a third would say.
- Citations aren't guaranteed
- Fixing a page makes it possible to be represented correctly. It doesn't make any platform cite it.
- Platforms differ
- Not every crawler reads pages the same way. Google-specific recommendations follow Google's published guidance; they aren't generalized to every platform.
- Not monitoring, not implementation
- Ongoing tracking and making the changes are separate scopes. The handoff gives your team what it needs to do both.
In their words
“Allison includes attributes in her AI Search assessment that I haven't seen on any other platform - she ties her analysis to the actual goals and objectives of the business. This moves beyond a basic technical scan and gets to what matter most to us.”
Excerpts from 2025-2026 company award submissions
Before you book
Common questions.
Is this SEO?
It overlaps. Access, clear content, and consistent entity information are ordinary SEO foundations. The difference is the observation set: we check what AI answers actually say about you, dated, and work back from there.
Our site is bigger than 25 pages.
The base tier reviews the 25 pages that matter most; we choose them together at kickoff. If you need broader coverage, the extended tier covers up to 75 pages, 50 customer questions, three answer platforms, and two competitor benchmarks. Beyond that, or across several markets, is quoted separately.
Will you rewrite the pages?
No. The report says exactly what to change and why. Your team or agency makes the changes. If the changes are content review at scale, the Content Resonance Engine is relevant.
Do we own the observation log?
Yes, along with the question set and the page review. Rerun it yourself in a quarter, or ask me to.
Bring the question a customer would ask about you. We'll see what an AI answer says, and work out whether a full assessment is worth it.
Free introductory conversation. No preparation required.
