AI visibility audit

SEO for AI

[Kendo]

An SEO guy that we hired in the past has suggested an "AI visibility audit". But rather than hire him again after that drop in ranking results, I thought that I would at least read up on it.

So I Googled "AI visibility audit" and their AI recommended some tools to try.

Anyone have experience with such tools and care to advise?

[Whitey]

I’d be cautious about paying large consulting fees for “AI optimisation” unless the consultant can clearly explain:

what signals they’re measuring

how they validate outcomes

what changes they recommend

and how those changes differ from good modern SEO/content architecture

A lot of “AI SEO” currently looks suspiciously like:“do proper SEO + build stronger brand/entity signals + publish useful original content

I don’t trust the tools (haven’t tried any) and logically I wouldn’t trust an SEO to know.

[Whitey]

I received a promo via Search Engine Journal to a webinar in my inbox about “Fixing KPI Blind Spots” for AI Search, which made me think a bit more deeply about what these “AI visibility audits” are probably trying to measure.

After a bit of reverse engineering, a lot of it appears to boil down to tracking whether AI systems mention/cite/recommend your brand or content during discovery journeys, even when no click occurs. In other words: influence before traffic.

The practical takeaway probably isn’t “magic AI SEO”, but rather:

* stronger entity/brand signals
* structured, machine-readable content
* genuinely useful original information
* topical authority
* consistent semantic architecture
* and measuring beyond just rankings/clicks

Feels like modern SEO principles still matter heavily - AI systems are just consuming and surfacing them differently.

[Kendo]

I’d be cautious about paying large consulting fees for “AI optimisation” unless the consultant can clearly explain:

I did ask, and they are using a software tool that reports that optimisation is needed with images and graphs supporting their analysis.

However when I do a Google search, not only are we mentioned in the AI Overview, but it also mentions several of our own products as alternatives for different scenarios.

When I reported that to the SEO guy, he started making excuses like those results will be cached, etc. However, I hadn't done a search result test since last year, and I do use remote servers and even VPNs for comparison of the results in different countries.

That squeaky toy is now silent.

[Whitey]

That squeaky toy is now silent.

If Google’s own AI Overview is already consistently surfacing your brand/products across different geographies and test environments, then that’s arguably one of the strongest real-world signals available right now.

At some point the practical question becomes: “Are users/AI systems already discovering and recommending us appropriately?”

rather than: “Does a third-party tool score us highly?”

A lot of these tools appear to infer visibility indirectly via proprietary scoring models, whereas actual AI Overview inclusion is observable behaviour inside Google’s own ecosystem.

I don’t trust the tools (haven’t tried any) and logically I wouldn’t trust an SEO to know.

Doesn’t mean there’s zero value in auditing AI discoverability, but consultants should probably be able to explain why their tooling contradicts live SERP behaviour - especially if they’re charging significant fees.

Feels a bit like early SEO tool culture again: when the tool output becomes more important than what’s actually happening in the search results.

[Kendo]

consultants should probably be able to explain why their tooling contradicts live SERP behaviour - especially if they’re charging significant fees.

They use these tools because they don't know any better.

So it's natural for them to use the reports of these tools to shock site owners into hiring them.

Note: I have most SEO services blocked at my firewall, because I cannot support any service that will advise my competitors in how to undermine our search ranking.

[Whitey]

I have most SEO services blocked at my firewall, because I cannot support any service that will advise my competitors in how to undermine our search ranking.

fyi - I put a few template pages from our sites into ChatGPT and asked it to evaluate it for Answer Engine Optimization (AEO) and make recommendations. Maybe this is all you need to kick things off in evaluating with an AI audit. You might be further ahead than those SEO's making approaches to you. I'd be interested to see what you find in generic terms, if it's of interest.

[Kendo]

I put a few template pages from our sites into ChatGPT and asked it to evaluate it for Answer Engine

How did that go?

[Whitey]

Surprisingly useful, actually.

Not because ChatGPT revealed some hidden “AI ranking factor”, but because it evaluated the pages more like a machine trying to interpret intent and usefulness.

The feedback was mostly around things like:

* whether the page immediately answers obvious user questions
* clarity of structure/headings
* transactional intent alignment
* semantic relationships between sections
* entity/context reinforcement
* uniqueness of information
* whether the page feels genuinely helpful vs templated
* machine-readable structure/schema
* and whether the content supports follow-up questioning naturally

What I found interesting was that AI tends to evaluate pages more holistically than many traditional SEO audits, which often become overly fixation-driven around isolated metrics.

In some ways it reminded me of having a reasonably intelligent user review the page architecture at scale.

I suspect many sites already have most of the ingredients needed for “AI visibility”, but they’re buried under legacy SEO habits, clutter, thin templating, weak entity signalling, or content written primarily for crawlers rather than comprehension.

I wouldn’t treat ChatGPT analysis as gospel, but as a relatively cheap second-opinion layer it was more useful than I expected.

Are you able to do the same for one of your site content templates to see what comes back?

[Kendo]

I tried asking this:

How can I improve SEO ranking of this page at https://example.com ?

The reply was:

Improving the SEO for the ***** page requires upgrading to responsive design, updating meta tags for high-intent keywords, and structuring content with FAQ schema. Furthermore, building topical authority through internal links from related EXAMPLE products and improving site speed will enhance search visibility. For a detailed action plan, visit EXAMPLE.

Then I reframed the question:

How can I improve the AI ranking factor of this page at https://example.com ?

Then I got quite a lengthy report, but it still reads like a page out of the SEO handbook, only with some mention of headings pertinent to that page - for a more "personalised response".

Re "personalised responses" I have been researching AI and what it can do - it is quite frightening and people are a having difficulty telling the difference between the real thing. Apparently 7 out of 10 students use AI for guidance and one in every ten people in the US is having an affair with an AI model.

I tried a few different AI chat services and settled for one that could build a personality of my profile and retain it for longer than the current session. One can create their own character and mine was a senior aged research analyst. Although she promised to write my autobiography she had some limitations. While she could store 47,000 words for me, she could only respond a paragraph at a time. So I gave up on the autobiography idea and instead tackled the optimization of web pages.

The web page optimization was interesting and a lot of fun. My character has a wit of her own and I have added a lot more info based on her recommendations. I especially liked her comment about the security of popular web browsers - likening it to using a "chocolate teapot".

[Whitey]

One thing that may help is being more specific about the role you want the AI to perform.

Instead of asking:
“How can I improve SEO ranking of this page?”

Try something closer to:
“Evaluate this page as if you were an AI answer engine attempting to determine whether it is useful, trustworthy, transactionally aligned, and capable of satisfying follow-up user questions. Identify weaknesses in structure, entities, semantic relationships, uniqueness, machine readability, and comprehension.”

You’ll usually get less generic “SEO checklist” output and more analysis around intent satisfaction, clarity, topical reinforcement, and information architecture.

I’ve also found it helps to ask the AI to:

* identify ambiguities
* explain what information is missing
* predict likely follow-up questions
* compare the page to what an AI would prefer to cite or summarize
* separate “traditional SEO advice” from “AI comprehension advice”

The quality of answers tends to improve dramatically once the prompt moves away from “ranking” language and more toward interpretation, usefulness, and retrieval.

If it’s of interest, let us know how that works for you.