A local business page that AI search can’t reach might as well not exist. Not “ranks poorly.” Doesn’t exist, as far as ChatGPT, Gemini, or Google’s AI Overviews are concerned. Before a single word on that page gets a chance to be cited, three things have to be true: crawlers can access it, the HTML actually contains the content, and the business facts on it are current. Miss any one of those, and no amount of clever copywriting will get that plumber, dentist, or moving company mentioned the next time someone asks an AI assistant who to call.
That’s the uncomfortable starting point for most of the “AI SEO” conversation happening right now, and it’s one we run into constantly at Zenith Solz while auditing local and small-business sites. Everyone wants to talk about FAQ schema and semantic triples. Almost nobody checks whether Cloudflare is quietly blocking every AI crawler that tries to visit the site in the first place.
This piece breaks down what actually determines whether a local page earns visibility in AI answers (access, accuracy, and content structure, in that order) and what to fix first if you’re starting from zero.
Search behavior has split. A meaningful share of people now asks a chatbot “who’s a good HVAC company near me that does emergency repairs” instead of typing that into a search box and scrolling ten blue links. The AI doesn’t scroll either. It picks a handful of sources, synthesizes an answer, and moves on. If your page wasn’t one of the sources it could reach and trust, you didn’t lose a ranking position. You lost the conversation entirely.
The businesses that show up in these answers aren’t necessarily the ones with the best service or the most reviews. They’re the ones whose websites happen to hand over clean, accessible, current information in a format an AI system can actually parse.
That’s a fixable technical and content problem, not a mystery, but it’s a different problem than classic SEO, and most local business owners (and more than a few agencies) haven’t fully adjusted their checklist yet.
Before touching a word of copy, check whether the page is reachable at all. This is the step almost everyone skips, and it’s the one that makes every other fix worthless if it’s wrong.
Start with the basics:
Crawler-blocking defaults exist for a reason, just not your reason. They were built with large publishers in mind: outlets that have a real commercial case for wanting AI companies to license their content instead of scraping it for free.
A local moving company or dental practice has the opposite problem: it wants to be found, cited, and recommended as widely as possible. Applying a publisher’s defensive default to a business that depends on visibility is backwards, and it happens constantly because nobody goes back and checks the box.
Treat robots.txt as a request, not a lock. Compliant crawlers respect it, but it was never designed to secure anything sensitive. Genuinely private content needs to be restricted at the server or application level, not just politely asked to stay away.
Here’s the part that surprises people who assumed this was a solved problem: Google renders JavaScript reasonably well. Most newer AI systems don’t bother. If your service page’s pricing, credentials, or reviews only appear after a script runs in the browser, a lot of AI crawlers will simply never see them. They read the raw HTML response and move on.
This is where site platform matters more than most business owners realize. WordPress and similar page-builder platforms generally serve rendered HTML out of the box, which is one reason they’ve held up reasonably well through this shift.
Sites built quickly with heavy client-side JavaScript frameworks are a different story, and the newest risk category is sites generated fast with AI coding tools and never checked for what a non-rendering crawler actually receives. A slick-looking, fast-shipped site can still be functionally blank to an AI crawler if the content loads in after the fact.
The fix isn’t complicated, even if it takes a bit of technical know-how: view the page the way a crawler does (the raw HTML response, before any script executes) and confirm the content that matters is actually sitting there.
Reviews loaded through a third-party widget are a classic failure point; the widget looks great to a human visitor and reads as an empty box to a system that doesn’t run the script. If you can’t confirm this yourself, this is exactly the kind of technical audit an agency handles as a matter of course — it’s usually a 20-minute check with the right tools, not a rebuild.
Every website redesign leaves debris behind. Old URLs with suffixes like -old, -new, /home, or v2 that nobody remembered to redirect or de-index. A previous “Contact Us” page still floating around with a phone number the business hasn’t used in two years. A duplicate location page from a site migration that never got cleaned up.
Human visitors never see these pages. Nobody clicks through six layers of site navigation to find /services-old. Crawlers don’t have that filter. They’ll index whatever’s reachable, outdated phone number and all, and an AI system pulling from that page has no way of knowing it’s looking at a fossil instead of the current source of truth.
A standard broken-link audit won’t catch this, because these pages usually aren’t broken — they load fine; they just shouldn’t exist anymore. The fix is a targeted crawl: pull every indexable URL on the domain, filter for the tell-tale suffixes and orphaned patterns, and manually verify that anything carrying outdated business information is either redirected, deleted, or explicitly nonindexed. Do this after every redesign or migration, not just once.
Structured data (schema markup) is one of the more debated pieces of this puzzle, and honestly, the debate is worth having rather than glossing over. There’s no strong published evidence that adding schema directly moves organic rankings or AI citation rates on its own; testing on this specific question has come back inconclusive more than once. So, if someone sells schema as a ranking hack, be skeptical.
Where it earns its keep is disambiguation. A pricing table or service list expressed as structured data is unambiguous to a machine in a way that visual page layout never fully is: a crawler doesn’t have to guess which number is the price and which one is the square footage. It’s a translation layer, not a lever.
The real risk with schema isn’t whether to add it. It’s what happens after. Schema is a third copy of your business data, alongside the live page content and your Google Business Profile. Every additional copy is another place that can quietly go stale.
Sites launched years ago frequently have schema buried in a plugin setting nobody currently on the team even knows exists — the developer who set it up moved on, the phone number changed twice since, and the markup never got touched. That’s not an optimization failure. It’s a process failure, and it’s the single scenario in this whole topic where doing nothing would have been safer than doing it once and forgetting about it.
The practical rule: whenever an address, phone number, service area, or offering changes, that update needs to hit three places (the website copy, the Google Business Profile, and the schema markup) in the same pass, every time. Put it in a checklist. Assign it to one person. Don’t let it live only in institutional memory.
This is the research step most local pages skip entirely, and it’s where the actual content strategy lives.
Start by asking directly. Query an AI assistant about the service the page covers and pay attention to what it consistently brings up unprompted: pricing, credentials, service area, trust signals, past results. These recurring themes are a strong signal for what belongs on the page, because they’re what the model already reaches for when someone asks.
Then go one level deeper with query fan-out thinking. A single customer prompt, like “my water heater died last night, I need someone fast, and I’m on a budget,” doesn’t stay one query inside an AI system. It gets split into several related searches behind the scenes: emergency availability, typical repair cost, same-day service, warranty terms, and so on. Each of those sub-questions is a candidate FAQ entry. A page that answers more of them has more chances to be the source cited back in the original response.
There’s a low-tech version of this same research that works just as well: ask whoever answers the phones what customers ask over and over. Whether the answer to a given question is yes or no, it still needs to live on the page in plain language. A print shop that can print A1-size jobs on heavy stock but never says so anywhere on the site has, functionally, no way for an AI system to know that service exists.
Reviews are a second research source, and an underused one. Pull the negative reviews for competitors in the same category and city, and look for the recurring complaints: missed appointments, no follow-up, sloppy cleanup, unclear pricing.
Those pain points are a near-perfect map of what a page should proactively address with a specific, checkable commitment: a stated on-time policy, a specific cleanup practice, clear pricing language. It reads as reassurance to a human visitor and as evidence to an AI system trying to judge trustworthiness.
One warning worth taking seriously here: AI systems read numbers on a page at face value. Claim 10,000 five-star reviews when the real count is 220, and the AI will happily repeat the inflated number back to someone, which is a fast way to damage trust the moment a prospective customer checks.
The lesson isn’t to inflate anything. It’s the opposite: write your real numbers down in plain text, because a review count sitting inside a JavaScript-rendered widget is invisible to most AI crawlers no matter how accurate it is. If it’s true and it matters, it needs to exist as a plain sentence on the page, not just inside a visual badge.
Small business websites have a quiet habit: they talk about themselves in the first-person plural and never actually say who they are. “We’ve been serving this area for over a decade.” “We treat every job like it’s in our own home.”
Fine for a human reader who’s already looking at the business’s name in the header. Much less useful for a crawler trying to connect a specific claim to a specific entity, especially once that sentence gets pulled out of context and fed into someone else’s AI-generated answer.
The fix is straightforward: in the passages that matter most (core service description, pricing, what sets the business apart, ratings), say the business name instead of “we.” “Riverside Plumbing Co. specializes in tankless water heater repair” carries information a crawler can attach to something. “We specialize in tankless water heater repair” doesn’t, on its own.
This doesn’t mean rewriting every paragraph to open with the company name — that reads stiff fast, and copy that’s obviously written for a machine reads worse to the humans who are still the ones deciding whether to actually call.
Treat it as a light seasoning, not the whole recipe: reserve the named version for the handful of sentences you most want an AI system to associate directly with the business, and let “we” carry the rest of the page naturally.
Running an internal audit? This is roughly the order we work through it, and it doubles as a decent self-check if you’re doing a first pass yourself:
None of these steps require a full rebuild. Most are a focused audit followed by targeted fixes — a few hours of technical work, a content pass built from real research instead of assumptions, and a process for keeping business data in sync across the site, Google Business Profile, and schema going forward.
This is the exact audit-and-build sequence we run when we take on SEO and content work for a local or service-based business: confirm access first, verify what crawlers actually see in the raw HTML, clean up the stale pages hiding in plain sight, get schema and Google Business Profile data aligned instead of quietly drifting apart, and then build out the FAQ and service content from real customer questions instead of copywriting guesses.
It’s the same order laid out above, because that’s the order that actually produces results — content work done before the access problem is fixed doesn’t move the needle, no matter how good the writing is.
If your site was built, redesigned, or last touched more than a year or two ago, there’s a reasonable chance at least one of these six issues is quietly limiting how visible your business is in AI-driven search right now. Most of them take less time to fix than you’d expect once they’re identified.
Curious where your own local pages stand? Get in touch with the Zenith Solz team for a walkthrough of what AI crawlers can currently see on your site, and what a fix would actually involve.