200+ AI tests reveal why some industries struggle with AI search

For 20 years, the web has operated with simple commerce: publish content that meets human needs, rank in search, gain traffic, and monetize that traffic through products, services, affiliate referrals, or ads.

Zero click responses and AI search are rewriting that relationship. The new question is whether AI will cite them as a source – and whether that can apparently turn into revenue.

To understand who is being included and who is being moved, I used more than 200 AI visibility tests across 10 industries.

The pattern was consistent: Most sites are easy to analyze, but hard to cite. And industries that rely heavily on traffic availability are often the ones that make it the most difficult to reach.

How the audit is done

I conducted 201 surveys using the same rubric and captured the visibility score of AI, along with four subscores:

• Burning.
• The structure.
• Authority and evidence.
• Detachable.

The database included 201 audits across 10 industries:

• Coupons.
• Related updates.
• Booking a trip.
• Location indicators.
• Comparison of personal finances.
• Health information.
• Legal documents.
• Online courses.
• Job boards.
• Recipes.

Note that there is page type skew — the sample is homepage heavy (131 pages, 13 articles, and a mix of pages left over). That’s important because home pages are often hard to sell and light proof.

I also tracked access failures because “error” results are part of the story. 38 out of 201 audits (18.9%) returned an error, meaning that the agent was probably blocked or could not reliably access the content.

Eight additional tests were technically evaluated but scored 0 due to a lack of young readers, accompanied by partial removal or app-style presentation that presented less accessible content.

When summarizing the distribution of points, I focused on audits that were successfully processed (163 sites), so that “I can’t access” was not combined with “low quality.” I treated the industry error rate as its signal because it showed that AI programs can reliably use the site as a resource.

Where industries stand in terms of AI visibility

The table below shows how the industries in the dataset performed in the audit.

Research findings revealed

The findings show that most websites are not always designed to be indexed. Here are three important numbers.

Access is a bigger problem than most groups think

38 of 201 sites (18.9%) returned an error. In other categories, it was worse: job boards (40%), legal directories (35%), travel bookings (33%), and course markets (30%). Of those spaces, a third to nearly a half of the market is AI-dark by default.

Legal documents have had the highest AI adoption of any industry.

Many sites are stuck in between

Among the 163 audits reviewed:

• Average overall score: 61.6
• Total score: 66
• 70.6% arrived at “inconsistent visibility” (60 to 79)
• Only 4.9% achieved “Strong Base” (80 to 94)
• 0% hit “Exceptional” (95 plus)

Translation: Many products are not designed to be used reliably and quoted.

A gap is proof, not formatting

Median support for all studies reviewed:

• Plot: 92
• Issue: 74
• Authority and evidence: 48
• Age: 45

Most pages are easy to analyze. Very few are easy to justify quoting. Two repeated results explain why:

• “No last modified header found” occurred 114 times (no machine-readable freshness).
• Citations or derivations appeared only 13 times (machine-readable proofs are rare).

That should change the way you think about risk. In addition to losing traffic, a major threat is removed from the consideration set.

Dig deeper: Revealing 4 AI tests on meaning and purchasing decisions

3 ways the industry is disappearing from AI search

Industries are disappearing for three reasons. You can think of them as three ways to fail.

1. Access failure: AI cannot reliably access your content

If agents can’t access your content regularly, the model has less to work with and will move around or fill in gaps from other sources.

What access failure looks like:

• Bot protection, rate limiting, or web application firewall (WAF) rules that treat agents as hostile.
• Application-style rendering where meaningful content does not arrive in the original HTML.
• Content is gated behind prompts, pirates, or scripts that don’t resolve cleanly.

Why this causes the disappearance:

• If AI systems can’t extract reliably, they can’t quote reliably.
• The user intent is still satisfied – they are simply satisfied with someone else’s crawling content or native AI response.

2. Failure of trust: AI can learn, but it cannot forgive you

The failure of trust is silent. An agent can access your page, analyze it, and summarize it, but the page doesn’t provide enough evidence for the model to confidently cite it as the source.

This was the dominant pattern in the completed audit. Simply put: Your content is readable, but not secure.

The clearest proof of this came when I compared page types:

• Average authority score for title pages: 76
• Median authority score for homepages: 45

A polished home page is not proof. If you want to be cited for anything more than your brand name, a generic homepage alone is not enough. Evidence usually resides in articles, descriptors, data pages, policy pages, and methodology pages.

3. Resource failure: Even if you are visible, clicks may not happen

Resource failure is very painful. You may be included. You may be cited. But if your value is only information, AI can compress it into a response, and the user doesn’t need to visit your site.

Visibility determines whether you appear in a conversation. The utility determines whether visibility turns into revenue.

A practical way to think about it:

• If your page answers a question, AI can change the page.
• If your product or service does the job, AI still needs you.

Failure to access will get you kicked out. A failure to trust makes you a failure. A resource failure causes you to be terminated.

Why certain industries are seen as vulnerable

When access, trust, and use are viewed together, the industries at risk cease to look random.

The categories that repeatedly showed the greatest vulnerability in my dataset shared three characteristics:

• Access is inconsistent (blocking and extraction problems).
• Content is easy to condense into one answer.
• The entity has no value for the next step once the response has been delivered.

That’s why travel bookings, job boards, legal directories, and coupon sites are grouped together as the most represented categories in this dataset.

The biggest takeaway? Your website can be designed in a way that invites exclusion, even if your business is healthy.

Dig deeper: Why each AI study tells a different story

A point not to be missed

Some industries will feel this harder than others. A site that is primarily funded by high volume informational traffic is highly exposed to zero click behavior. But even in those categories, the way forward is to stop selling information on your own.

The biggest mistake right now is treating AI search as a ranking update, when it’s an economic update. The audit made two things clear:

• Many factories make it difficult to reach them, which ensures that the model will move around them.
• Even if the model can read the page, it usually cannot justify the claim because the evidence is not there.

The threat is invisible. You don’t win by hiding. You win by being quotable and creating something the user still needs after the response is delivered.

Trust plus utility is a new route. Anything else is just playing from yesterday’s playbook.