When AI Search Ignores Your Small Business: What We're Seeing in Real Audits
We built EVOIX's AI Readiness Audit to answer one question every small business owner should be asking: when a customer goes to ChatGPT, Claude, or Gemini, what's actually happening with your business in those answers?
We're still early, this post covers observations from the first cohort of sites we've audited, not a large-sample study. The patterns we're seeing so far are consistent enough, and the fixes are concrete enough, that they're worth sharing now. We'll publish a larger dataset once we have one.
What the Audit Actually Does
Before any findings, it helps to know what we measured. Every audit runs two things against a site:
A technical scan of the website itself, we check for structured data (JSON-LD schema), the quality of meta tags, the presence and structure of an FAQ section, the depth of the About/Company information, Name-Address-Phone (NAP) consistency, heading structure, content depth, robots.txt, sitemap, and a few experimental signals like llms.txt.
A set of queries against the major AI platforms, we send 15 prompts across OpenAI, Claude, and Google Gemini. The queries split into three types: brand queries (asking the AI about the business by name), category queries (asking about the top businesses in the industry), and recommendation queries (asking the AI to recommend someone for a specific service).
Two caveats upfront. First, our current API queries don't enable grounded web search, so they reflect what base models "know" about a business from training data, not what a real customer would see in a grounded ChatGPT conversation. The technical-scan half of the audit is where our highest-confidence findings sit today. Second, our sample is early and self-selected, business owners who ran the audit on their own sites. We're not making statistical claims about the small-business market as a whole. We're reporting what we're seeing on the actual sites we've scanned.
The Technical Signal Patterns
This is the part we can speak to with real confidence. Every site we've scanned gets graded on the same technical checklist, and the gaps are showing up in the same places.
Structured data is missing on a meaningful share of sites. Roughly one in four sites we've audited has no JSON-LD schema markup at all. Schema is the single highest-impact change most small businesses can make for AI visibility, it's a small block of structured code in the <head> of your site that tells AI engines exactly what your business does, where you operate, and what services you offer. Without it, AI has to parse unstructured marketing copy and guess.
FAQ sections are often thin or absent. Around one in four sites had no dedicated FAQ content. AI engines love structured question-and-answer content because it matches exactly what an AI engine is trying to produce, an answer to a question. A page with ten real customer questions and ten direct answers is dramatically easier for an AI to extract and cite than a 2,000-word paragraph essay.
NAP consistency averages roughly 70% on the sites we've scanned. That means on most sites, the business name, address, or phone number shows up slightly differently across footer, contact page, Google Business Profile, and directory listings. AI engines use NAP consistency as a trust signal. Small inconsistencies, a suite number missing here, a phone number formatted differently there, tell the AI the data source isn't canonical.
About pages are often thin. On the sites we've scanned, About pages frequently lack founder names, photos, founding year, or specific achievements. AI engines look for "entity signals" that a business is a real operation with a real history. An About page that says "We're passionate about serving our community" gives them nothing to cite.
The pattern across these findings is the same: none of them are hard problems. Schema takes an afternoon. A real FAQ takes a couple of hours. NAP cleanup is mechanical. Filling out an About page is writing, not engineering. These are the lowest-hanging fruit in AEO, and most small-business sites we see are leaving them on the ground.
What We're Still Learning About AI Mention Rates
The other half of the audit, the AI platform queries, is where the dataset needs to grow before we draw conclusions. We've seen the expected pattern so far: AI platforms can often locate a business when asked by exact name, but rarely surface it in generic discovery queries. That's consistent with what AEO researchers have been saying for over a year.
But with a small, self-selected sample and ungrounded API queries, we're not ready to publish mention-rate percentages. We're adding grounded-search queries to the audit next, and we'll revisit this data publicly once we have a materially larger cohort and the methodology to match. When that post goes up, you'll be able to see it at the same URL with updated numbers.
Five Fixes Worth Doing This Week
Based on the technical signals the audits are surfacing, here's where the time is best spent:
1. Add LocalBusiness schema to your homepage. Google's Schema Markup Generator will produce valid JSON-LD from a short form. Paste it into your site's <head>. One afternoon, durable result.
2. Build a real FAQ section with 8-15 questions. Use the questions customers actually ask on sales calls. Pricing, turnaround, service area, guarantees, what's included. Answer each in 1-3 sentences. Use <h3> for the questions so the structure is machine-readable.
3. Make NAP consistent everywhere. Audit your Google Business Profile, footer, contact page, social profiles, and directory listings. Pick one canonical format and enforce it.
4. Expand your About page. Include the founder's name, a real photo, a concrete founding story, specific years, specific locations. AI engines want to cite businesses that look like businesses.
5. Publish content structured as direct answers to direct questions. Title posts with the exact customer question. Lead with the answer. The older listicle format ranks worse in AI contexts than a clean "question to direct answer to supporting detail" structure.
Where to Start
If you've never checked how your site looks through the lens an AI engine actually uses, the audit is free and it gives you a prioritized list of fixes specific to your site. Take our free AI Readiness Audit here.
We'll be publishing updated findings as the dataset grows and the methodology tightens. In the meantime, the five fixes above are the ones we'd recommend to any small business owner who wants to start showing up when AI answers the questions their customers are asking.