SEO »
An audit of 71 businesses found major gaps in AI retrievability, from unreadable websites and dead domains to missing trust signals. #
There’s a gap between what a business is in the real world and how an AI search system can verify it. I call this the “identity leak.”
Businesses may not be struggling, but they may be invisible to AI systems, which decide what sites get recommended and surfaced to searchers. This can affect a business’s future growth.
I ran an audit of dozens of verified businesses on Prince Edward Island (PEI), Canada, spanning multiple industries, to see how they showed up in AI search results.
The audit found that the average PEI business was leaking 84% of its identity to AI systems, and 17% of the sample had no AI-retrievable digital presence at all.
Here’s what the identity leak is, why it happens, and how to close it.
What is the 84% identity leak? #
An AI search retrieval system doesn’t browse a website the way a person does. It looks for specific, verifiable facts: who runs a certain business, where it is, what it does, whether the people behind it are real, etc.
When it can’t find these, it guesses, using whatever fragments it can piece together from the web. Or it’ll simply leave the business out of an answer entirely.
The gap between what’s true about a business and what an AI can confirm about it is the identity leak.
Here are some examples of how businesses appeared in the audit I conducted.
Trust signals exist, but AI can’t surface them
This is the most common and most fixable version of the leak. Twenty-two of the 71 businesses in the sample had named, identifiable leadership somewhere on their own site.
In several cases, that information lived on an “Our Team,” “Our History,” or “Our Family” subpage that a routine pass of the homepage by an AI missed.
The information is there, it’s just not where the verification needs to be.
[ Be the brand AI recommends. See your AI visibility
](https://www.semrush.com/ai-seo/overview?utm_campaign=ic_sel_0101ai&utm_source=searchengineland.com&utm_medium=overlay&onboarding=off) See where your brand appears in AI search, where competitors are winning, and what it takes to become the answer AI recommends.
Site exists, but nothing is readable by an AI
Several businesses in the sample had professionally built, real, modern-looking websites that returned zero extractable text to a direct fetch.
They were built entirely in client-side JavaScript with no static fallback. One of these was a software company whose own tagline promised “intuitive enterprise and AI systems.” Its homepage was unreadable to the very system it claimed to be built for.
Business is real, but the domain is dead
One well-known cheesemaker’s domain is now for sale by a domain reseller, and a salt producer’s domain returned nothing. The brand now survives only as a product line sold through another website. Neither business showed up where a crawler was looking.
Identity is split across the web
One chocolatier circulated under three different domain variants across directories, and a biotech company operated two separate live domains for the same entity.
Each of these fragments dilutes the others, and an AI system will try to resolve “who is this,” finding competing, partial answers instead of one authoritative answer.
Business never built a digital presence
A handful of real, operating businesses in the sample, including a photography studio, an HVAC contractor, a lumber yard, and an auto shop on the province’s own vehicle-inspection registry, exist only in third-party directory listings. They don’t have actual sites for an AI system to verify against.
Dig deeper: The new playbook for localized AI search optimization
Why the identity leak happens #
The identity leak isn’t the result of bad business. It’s the result of websites that were built for a different, older kind of reader.
In the past, a website’s job was to be found by a search engine and read by a human who already had some context: a brand from an ad, a recommendation from a neighbor, a listing in a local directory, etc. Under those conditions, a website that looked good and loaded fast was doing its job, and AI readability wasn’t a consideration.
AI systems and how they retrieve information have changed the game. A system that extracts facts about a business needs structured, current, and unambiguous data.
Three technical issues explain most instances of the leak found in this audit.
Client-side rendering with no static fallback
If a page’s content only exists after JavaScript executes in a browser, a retrieval agent that parses static HTML will never see it. This produced the worst results in the study: no content at all.
Real facts confined to pages nobody checked
Named leadership, real history, and policy pages disproportionately lived on secondary pages rather than the homepage.
A full re-verification pass caught most of these errors, evidence of how easy this gap is to miss and how real it is once you actually look.
No canonical source of truth
In instances where a business’s identity is split across multiple domains or directory listings, or a dead canonical domain exists alongside a living social presence, an AI system has no way to determine which source to trust.
It often defaults to whichever third-party platform has invested the most in making that site verifiable. In hotel and golf-resort accommodations specifically, this showed up as third-party booking resellers that ranked alongside or above the property’s own booking page in search results — a version of the leak with a direct commission attached.
Dig deeper: The perfect local business contact page built for Google and conversions
How the identity leak audit worked #
The structured audit was conducted on 71 verified businesses across food and beverage, retail, professional services, technology, agriculture, health care, accommodation, and golf.
I used a condensed version of an E-E-A-T scoring framework, built around the same principles search engines have used for years to evaluate quality and whether a page deserves to rank. The framework was modified for how AI retrieval systems trust and extract information now, using a points-based diagnostic.
The full audit scored businesses across five categories, totaling 500 points:
- Main company entity audit (mission, leadership visibility, history, contact information)
- Technical foundations (HTTPS, legal pages)
- Initial data collection (backlink quality, social signals, content recency)
- Senior entity
- Policy page
The condensed version scored businesses against a fixed 485-point scale — the 500-point baseline minus 15 points reserved for Core Web Vitals, which requires tooling that this audit didn’t have access to. All businesses were scored against the same standard.
In addition to the E-E-A-T score, other checks included:
- If a business’s name, address, and phone number were consistent across its own site and the sources that reference it.
- If a real, named human being was identifiable on the business’s own domain.
- If a direct fetch of the site returned readable text.
This low-level AI retrievability check was designed so anyone could run it. It’s not a deep technical audit of crawl budgets, structured data implementation, or server infrastructure. None of these checks require specialized tooling. A person with no technical background could perform every one of them by hand, with a browser, in under five minutes per business.
Results were independently rechecked in a full second pass after the initial findings. Every business, statistic, and external source cited is independently verifiable, with a website link for each of the 71 businesses and a complete source list with live citations to Statistics Canada, CoStar/STR, Cloudbeds, the National Allied Golf Associations, and other primary sources (the full underlying study, including a complete source list, is available on request).
My audit found that the average business resolves only 15.6% of what it needs to be considered trustworthy and retrievable for an AI. Or, it’s leaking about 84% of its identity.
If a basic check can find a gap this large, a business isn’t falling behind because of a technical issue, but because nobody has looked at the problem. These are simple fixes — changing a named owner on an existing page, a linked privacy policy, or a confirmed live domain. Dig deeper: The local SEO gatekeeper: How Google defines your entity
How to identify and close the identity leak #
In nearly every case in this audit, the same handful of interventions closed the gap. Here’s how to do it.
Put a named person on the page: The most common fix across this study was to add a real “About the Owner,” “Our Story,” or “Our Team” page with an actual name. If this already exists on a secondary page, make sure it’s linked to the homepage.Server-render content that matters: Any business that relies on a fully client-rendered site needs a static fallback for its core facts (name, address, services, leadership), even if the interactive experience stays as-is for human visitors.Pick one domain and commit to it: Where a business operates across fragmented or duplicate domains, consolidate to a single canonical address and redirect everything else to it.Confirm the domain is actually live: This sounds basic, but it wasn’t for at least two businesses in this sample, whose domains had quietly lapsed or been listed for resale without anyone noticing.Add a real, linked policy page: Privacy policies and terms of service are trust signals that AI systems weigh, and they weren’t always present, even among well-built sites.Check whether a third-party reseller is ranking near or above your own booking page: For businesses with a transactional booking path, e.g., hotels or anything where an AI system might route a paying customer, make sure your direct path is at least as verifiable as theirs.
[
If AI can’t find you, customers won’t either.
See your AI visibility
](https://www.semrush.com/ai-seo/overview?utm_campaign=ic_sel_0102ai&utm_source=searchengineland.com&utm_medium=overlay&onboarding=off) Track your visibility across AI search, uncover missed opportunities, and grow your presence where customers are asking questions.
Make your business visible in AI search today #
The identity leak isn’t a story about bad businesses or lazy marketing. Every business named in this research is real, established, and in most cases, excellent at what it does.
The identity leak happens when businesses are represented online by infrastructure that was built for an old technology, when human readers already had a good sense of where to look.
The businesses that scored best in this study weren’t the most resourceful or the most aggressive marketers. They were the ones with an institutional obligation to be transparent and therefore had to make their sites easily readable to systems: a regulated care facility, a publicly accountable nonprofit, and a government-marketed asset with a dedicated general manager, for example.
Closing the identity leak isn’t expensive, and it isn’t hard. It’s mostly unknown and just a matter of businesses knowing how to look for it.
Contributing authors are invited to create content for Search Engine Land and are chosen for their expertise and contribution to the search community. Our contributors work under the oversight of the editorial staff and contributions are checked for quality and relevance to our readers. Search Engine Land is owned by Semrush. Contributor was not asked to make any direct or indirect mentions of Semrush. The opinions they express are their own.