77%

Last updated: March 2026

We checked 30 UK small business websites for signals that could cause confusion when AI systems try to understand what the business does. The majority had at least one pattern that introduces ambiguity.

---

TL;DR

• 77% of sites use homepage carousels or sliders — rotating content that presents multiple messages where AI systems expect one clear signal
• 50% have no H1 heading — the primary heading that tells machines what the page is about
• 30% have no meta description — leaving AI with no page summary to work with
• 20% mention reviews without linking to a source — unverifiable trust claims
• 20% show signs of stock photography rather than original imagery
• When these signals combine, they create a homepage that sends unclear or contradictory signals about what the business is

---

What makes a website confusing to AI?

AI systems process websites differently from human visitors. A human sees the full visual experience — layout, imagery, colour, animation — and processes it instantly. They understand that a carousel is showing multiple messages, that a stock photo represents a concept rather than the actual team, and that "YOUR SUCCESS STARTS HERE" is a motivational slogan rather than a description of the business.

AI systems parsing HTML do not have this contextual processing. They encounter:

• Multiple content blocks within a carousel, potentially with conflicting messages
• An H1 heading that says something different from the meta description
• Review claims with no verifiable source
• Generic imagery described by alt text that may not match the business
• Motivational or creative copy that does not describe the business type

None of these are "wrong" in web design terms. They are standard practices used by millions of websites. But for AI systems trying to determine what a business does, who runs it, and whether it is credible, these patterns introduce noise.

---

What we found

Signal confusion patterns

Pattern	Count	Percentage
Carousel or slider on homepage	23 / 30	77%
No H1 heading	15 / 30	50%
H1 does not state what business does	16 / 30	53%
No meta description	9 / 30	30%
Mentions reviews without linking to source	6 / 30	20%
Stock photo indicators in HTML	6 / 30	20%

The carousel problem

77% of the sites in our sample used some form of carousel, slider or rotating content on their homepage. This is an extremely common web design pattern — and it is not inherently problematic for human visitors who see each slide in sequence.

For AI systems, carousels present a specific challenge. The HTML typically contains all slides' content simultaneously, meaning the AI encounters multiple headings, multiple messages and potentially multiple calls to action within the same page section. If slide one says "Award-Winning Accountants" and slide three says "Transform Your Tax Strategy", the AI is processing both messages as concurrent homepage content.

This does not mean carousels prevent AI understanding. But they add content noise — multiple messages where a static page would present one clear signal. For businesses already struggling with entity clarity (generic names, missing schema, no H1), a carousel adds another layer of ambiguity.

The compounding effect

The most concerning finding was not any individual signal but how they combine. Consider a site that has:

• No H1 heading (AI has no primary heading signal)
• No meta description (AI has no page summary)
• A carousel with three different messages
• No Organisation schema
• Review mentions with no links

This site is asking an AI system to determine what the business does based on body text alone, filtered through multiple competing messages, with no structured data to anchor the interpretation. It is not that the business is invisible — it is that the signals are noisy, and a noisy signal is harder to interpret accurately.

How many sites had multiple confusion signals?

Confusion signals present	Count	Percentage
0-1 confusion signals	4 / 30	13%
2 confusion signals	8 / 30	27%
3 confusion signals	11 / 30	37%
4+ confusion signals	7 / 30	23%

Only 13% of sites had one or zero confusion signals. Nearly a quarter had four or more. The typical UK SME homepage in our sample had 2-3 patterns that could introduce ambiguity for AI interpretation.

---

Real examples from our check

Example 1: The estate agent with a blog post as its H1

One estate agent — H1: "Signs You Might Be Ready to Move (Even If You're Not 100% Sure)"

This H1 reads like a blog post title, not a business identity statement. Combined with no Organisation schema, no Person schema and no sameAs links, an AI system encounters a page that could belong to a moving company, a self-help blog, a relocation service or an estate agent. The actual business type must be inferred from secondary content.

Example 2: The plumber that doesn't say "plumber"

One plumbing company — H1: "London's No.1 - improving and maintaining properties for over 45 years"

The H1 claims market position ("No.1") and describes what they do generically ("improving and maintaining properties") but never uses the word "plumber", "plumbing" or "heating". An AI system processing this H1 in isolation might classify it as a property maintenance company, a building contractor, or a facilities management firm.

Example 3: The accountancy firm with a provocation instead of a description

One accountancy firm — H1: "It's The End For Accounting Dinosaurs..."

Creative and attention-grabbing for humans. For an AI system, this headline discusses the end of something and dinosaurs. The word "accounting" is present but in a metaphorical context. The actual service (accountancy) must be extracted from the creative framing.

Example 4: The personal trainer with a generic motivational statement

One personal trainer — H1: "YOUR SUCCESS STARTS HERE"

This heading could apply to virtually any business in any industry. An AI system reading this has zero signal about the business type. In this case, the domain name provides more entity information than the H1 itself.

---

The stock photography signal

20% of sites showed indicators of stock photography in their HTML source — references to Shutterstock, iStock, Getty, Unsplash or Pexels in image URLs, alt text or filenames.

This is notable not because stock photography is bad — it is a normal, practical choice for many businesses — but because it creates a disconnect between the imagery and the business entity. When alt text describes a stock photo ("happy team meeting in modern office") rather than the actual business ("the Summit Digital team at their Manchester office"), the image metadata provides generic rather than entity-specific signals.

For AI systems that process image alt text as part of their entity understanding, stock photo descriptions add noise rather than clarity. They tell the AI what the image depicts generically, not what it means in the context of this specific business.

---

What businesses can consider

This is not about redesigning websites. Most of the patterns we found are standard web design practices that work perfectly well for human visitors. The question is whether small adjustments could reduce ambiguity for AI systems without compromising the human experience.

Quick clarity improvements:

1. Make the H1 descriptive. If your H1 is a slogan, consider making it a clear statement of what you do. You can keep the slogan as a subheading.
2. Add a meta description. 30% of sites had none. This is a one-line addition that provides a page summary for both search engines and AI systems.
3. Keep the carousel but lead with your clearest slide. If slide one clearly states what the business does, the AI encounters the right signal first even if subsequent slides add complexity.
4. Link review mentions to their source. If you mention a rating, link to where it can be verified.
5. Use original alt text. If you use stock photos, at least write alt text that describes the photo in your business context rather than using the generic stock description.

These are small changes. None of them require a website rebuild. They reduce signal noise without changing the visitor experience.

---

Methodology

• Sample: 30 UK SME websites selected from web search results across nine industries
• Date: 12 March 2026
• Carousel detection: HTML scanned for keywords indicating carousel/slider libraries: "carousel", "slider", "swiper", "slick"
• Stock photo detection: HTML scanned for references to known stock photo platforms: "shutterstock", "istock", "getty", "unsplash", "pexels"
• H1 and meta description: Extracted from homepage HTML source
• Confusion signal count: Sum of: no H1, unclear H1, no meta description, carousel present, unlinked review mentions, stock photo indicators. Maximum possible: 6
• Limitations: Carousel detection may produce false positives (some sites may reference these terms without using carousels). Stock photo detection only catches references in HTML source — some stock images may have had their source references removed. These checks assess individual signals, not how AI systems actually process the combined page. This is a convenience sample and should not be extrapolated to all UK businesses.

---

FAQ

Do carousels hurt AI visibility?

There is no confirmed evidence that carousels directly reduce AI visibility. However, carousels present multiple messages simultaneously in the HTML, which may introduce content ambiguity for AI systems trying to determine the primary purpose of a page. A static hero section with a single clear message is unambiguous in a way that a carousel with three different messages is not.

Should I remove my carousel?

Not necessarily. Carousels work well for human visitors and are a standard design choice. If you keep your carousel, consider ensuring that the first slide clearly states what the business does, and that the H1 heading (which should be outside or separate from the carousel content) provides a clear, static identity signal.

Does stock photography confuse AI?

The images themselves are not the issue — AI systems do not typically evaluate image quality or originality. The concern is with alt text and image metadata. If alt text describes a generic stock photo rather than something specific to your business, it provides generic rather than entity-specific signals. Writing custom alt text for stock images is a simple way to address this.

How many confusion signals is too many?

There is no established threshold. Our observation is that businesses with multiple confusion signals (missing H1, no meta description, carousel with mixed messages, no schema, unlinked reviews) present a noisier signal profile than those with fewer. Reducing any one of these signals improves clarity marginally; reducing several improves it substantially.

What's the quickest fix for a confusing homepage?

Add or update the H1 to clearly state what the business does. This is typically a single text change and provides the strongest early signal to AI systems. If the H1 already exists but is a slogan, consider changing it to "[Service type] in [Location]" and moving the slogan to a subheading.

Can AI handle confusing websites?

Yes. AI systems are sophisticated and can often determine what a business does even from noisy signals. The question is not whether AI can handle confusion — it is whether a clearer signal profile reduces the chance of misinterpretation. For businesses in competitive categories where AI platforms are choosing between similar businesses to recommend, signal clarity may be a marginal advantage.

Is this only about AI or does it affect traditional SEO too?

Many of the same principles apply. Missing H1 tags, absent meta descriptions and unclear homepage messaging are traditional SEO issues as well as potential AI visibility issues. Addressing them improves clarity for all systems that process the page — search engines, AI platforms and human visitors alike.

What does a "clean" homepage look like from an AI perspective?

Based on our research: a clear H1 stating what the business does, a meta description summarising the business, Organisation schema with sameAs links, a static hero section (or a carousel with a clear first slide), review mentions linked to verifiable sources, and original or well-described imagery. Only a small minority of the sites we checked had all of these in place.

---

This research was conducted by Rank4AI as part of our ongoing work understanding how UK businesses appear in AI-powered search platforms. Our findings are observational and should not be taken as guarantees of specific outcomes.

For more on how AI interprets business websites, see our guides on why AI misinterprets businesses, meaning architecture, and diagnosing weak AI meaning signals.

Want to understand how these trends affect your business?

Start Your AI Visibility Review