We spent a week this month running 20 small business homepages — a mix of client sites, prospect sites, and businesses we picked at random from local directories — through the same checklist we use to judge whether a page is legible to AI answer engines. Only three passed on the first read. The other seventeen shared a small handful of the same five problems, repeated over and over regardless of industry. If you want to know where to spend your next hour of homepage work, this is probably it.
None of these businesses had bad websites in the traditional sense. Most looked fine, loaded fast enough, and would pass a normal design review. The gap was almost entirely about whether an AI system reading the page could extract a clean, confident answer to “what does this business do, for whom, and why should I trust them” — the exact question ChatGPT, Perplexity, and Google’s AI Overviews are trying to answer on a searcher’s behalf.
What are the 5 fixes that came up again and again?
In order of how often we found them, across the 20 sites: no clear one-sentence description of the business above the fold, no author or ownership information anywhere on the site, mixed or missing heading structure, service descriptions written for humans skimming rather than for extraction, and zero specific numbers anywhere on the page.
1. No clear one-sentence description above the fold (17 of 20 sites)
This was the single biggest offender. Most homepages opened with a tagline built for brand feel rather than clarity — phrases like “Building the future of X” or “Your partner in growth” that mean almost nothing to a person or a language model scanning the page. When an AI system tries to summarize what a business does, it needs a direct sentence to lift, and if the page doesn’t provide one, the model either guesses from secondary signals or skips the business in its answer entirely.
The fix is not clever copywriting — it’s the opposite. State plainly, in the first 200 words of the homepage, what the business does, who it serves, and where. “We design and install solar panel systems for homeowners in Southern Finland” beats a mood-board tagline every time a model is deciding whether to cite you.
2. No author or ownership information (16 of 20 sites)
Sixteen of the twenty sites had no visible answer to “who runs this business.” No named founder, no team page, no author bylines on blog content, sometimes not even a physical address beyond a contact form. This matters more than most business owners assume, because trust signals are exactly what E-E-A-T-style evaluation — used by both traditional search and AI answer engines — is built to detect. We wrote more on how to build these signals without a big brand name behind you in our piece on E-E-A-T for small businesses, but the short version is: name a real person, put a face and a short bio somewhere findable, and don’t hide behind a generic “info@” contact form as your only trust signal.
3. Mixed or missing heading structure (14 of 20 sites)
This one is invisible to a human visitor and very visible to a machine reading the page’s HTML. Several sites used bold, larger text to look like a heading without using an actual heading tag, or skipped straight from an H1 to an H4 with nothing in between, or used the same H2 six times for visually similar but semantically different sections. AI systems and search crawlers alike use heading hierarchy to understand the shape of a page’s argument — what’s a main topic, what’s a subtopic, what’s a supporting detail. A flat or scrambled hierarchy makes that shape unreadable, even when the content underneath is genuinely good.
The fix here is mechanical, not creative: one H1 per page, H2s for major sections, H3s nested logically underneath. If you’re not sure your site does this correctly, open the browser’s “Inspect” tool and search the HTML for <h1>, <h2>, <h3> — you’ll see the real structure in about ten seconds.
4. Service descriptions written for skimming, not extraction (13 of 20 sites)
Most service sections were built to be skimmed visually — an icon, a three-word label, a single vague sentence, repeated across a grid of six services. That layout works for a human eye moving quickly down a page. It works badly for a model trying to extract “what exactly does this service include” as a discrete, quotable fact. The service blocks that performed best in our read-through had at least one full sentence per service that could stand alone, out of context, and still make sense: what it is, who it’s for, and roughly what’s included.
5. Zero specific numbers anywhere on the page (12 of 20 sites)
Years in business, number of clients served, a size range, a typical turnaround time, a starting price band — something concrete. Twelve of the twenty sites had no numbers on the homepage at all beyond a phone number. Specific, quotable numbers are exactly what a model favors when constructing a citation, because they read as verifiable rather than promotional. “We’ve built websites since 2016” is more citable than “we’ve been in the industry for years,” and it costs nothing to write.
How does this connect to your existing content, not just the homepage?
The homepage carries more weight than any other page for this kind of read because it’s usually the page a search or AI system encounters first when trying to understand what a business is. But the same five patterns showed up on interior pages too, and the fixes compound: a site with clean heading structure and strong internal linking gives a model a much easier path from the homepage into the pages that actually answer a searcher’s specific question. If your internal linking is thin, even a strong homepage won’t carry the rest of the site — we go deeper on that in our internal linking strategy for AI-era SEO.
It’s also worth checking whether your service pages already have the structural signals search engines and AI systems reward, which is the same territory we cover in our answer-engine optimisation checklist and, for local businesses specifically, in how structured data helps local pages appear in AI Overviews.
What would we do first if this were our own site?
If you only have an hour this week, spend it on fix #1 and fix #5 together: rewrite your homepage’s opening sentence to state plainly what you do and for whom, and add two or three specific numbers into that same section — a founding year, a client count, a turnaround time. Those two changes take under an hour combined, require no design work, and address the two most common gaps we found across all twenty sites.
The heading structure and author information fixes take longer because they touch more of the site, but they’re mechanical enough to hand to a developer or virtual assistant with a checklist rather than needing a full rewrite. None of the twenty sites needed a redesign to fix what we found. They needed about half a day of focused editing, aimed at the handful of things that were actually costing them visibility.
The takeaway
Small business homepages tend to fail AI search readiness for structural and clarity reasons, not content-quality reasons. The businesses in our sample of twenty had real expertise and legitimate things to say — the pages just weren’t built to let a model extract that expertise cleanly. Fixing that is closer to copyediting and HTML hygiene than it is to a content strategy overhaul, which is good news: it’s genuinely a one-afternoon project, not a redesign.