More Than 4 in 10 Small Business Websites Is Accidentally Telling Google It's About a Person

A study of 85 SMB homepages reveals a widespread — and almost entirely invisible — entity classification problem that may be suppressing local search visibility.

There is a word sitting near the top of thousands of small business websites right now. It is not a proper name. It is not a job title. It is not anything a human reader would flag as unusual.

Google's entity analysis is classifying it as a person.



That misclassification — quiet, automatic, invisible to anyone without access to entity-level analysis tools — is likely suppressing the local search visibility of the website it lives on. And based on a study of 85 valid small business homepages conducted across the construction, trades, and functional medicine verticals, it happens on more than 4 in 10 of them.

When Google processes a webpage, it doesn't just read words — it attempts to understand what the page is about. One of the mechanisms it uses is entity analysis: identifying the people, places, organizations, and concepts on the page and assigning each one a salience score, which represents how central that entity is to the overall meaning of the document.

Salience is a finite budget. The scores across all entities on a page sum to roughly 1.0. When one entity captures more, the others capture less.

For a local business, this has a direct implication: the geographic entity (a city, a region, a market) needs to be well-represented in the top salience positions for the page to clearly signal local relevance. So does the organization or service category. When something else — especially a person — claims an outsized share of that budget, the signals that matter for local search can get crowded out entirely.



This is what the study set out to measure.

The PERSON Entity Trap occurs when a webpage contains language that causes Google's entity analysis to classify a non-person entity as a PERSON type — triggering a salience draw that the site's owner never intended.

There are two distinct mechanisms:

The Physician-as-Brand mechanism occurs when a solo practitioner or founder's name is embedded throughout a homepage — in the domain, the title tag, the H1, and the body copy. The site intends to personal-brand around that individual. The unintended consequence is that the person consumes the majority of the page's entity budget, leaving little room for service and location signals.

The most extreme example in this study: a functional medicine practice whose homepage assigned 81.7% of its total entity salience to the founding physician's name — leaving a combined 18.3% for every other signal on the page, including the city it serves.

The Generic Language mechanism is the more surprising finding. It occurs when ordinary, unremarkable copy — the kind that appears on thousands of websites — contains words that entity analysis classifies as PERSON type.

Examples confirmed across the study:



• "our people" and "put our people first"
• "our patients saw improvement"
• "builders who care about craft"
• "a team of experts"
• "employees are our strongest resource"
• "clinicians who listen"
• "you are the central partner in your healing"
• "whether you are an adult or child"
• "We Are Builders" (headline format)

None of these phrases would raise a flag in a copywriting review. They are the standard vocabulary of service business marketing copy. But entity analysis does not read them the way a human does — it reads them the way a classifier does. And a classifier trained on patterns across the entire web has learned that "patients," "builders," "clinicians," and "employees" are most often references to people.

The study analyzed the homepages of 85 valid small business websites across the construction, trades, and functional medicine sectors. Sites that returned no readable body copy — due to JavaScript rendering, maintenance pages, or spam injection — were excluded from the count and replaced where possible. All analysis was performed on homepage copy only, using homepage as the consistent comparison point across all sites.

Finding 1: More Than 4 in 10 Sites Has a PERSON Entity in Its Top Three

Across the full dataset, 41% of sites had a PERSON entity classified in their top three entity signals — in cases where that entity was not, in fact, a named person intentionally featured on the page.

In plain terms: more than 4 in 10 small business homepages is sending a signal that the page is, in some meaningful part, about a person — when the business intends to signal a service, a location, or an organization.

Finding 2: In Most of Those Cases, the Location Signal Disappears Entirely

This is the mechanism that connects the PERSON Entity Trap to search visibility outcomes.

Of the sites where a PERSON entity appeared in the top three, 63% had no location entity in their top three signals at all. The geographic entity — the city or region the business serves — had been pushed out of the top positions entirely.

This is what the study terms Salience Budget Collapse: the condition where PERSON entity misclassification consumes enough of the page's entity budget that the local relevance signal fails to surface in the top results of entity analysis.

For a local business that depends on geographic search visibility, this is a structural problem. The page is being understood by entity classifiers as being about a person or a group of people — not a roofing company in Dallas, not an integrative medicine practice in Chicago, not a plumber in Denver.

Finding 3: Two Distinct Mechanisms Drive the Trap

The PERSON Entity Trap does not have a single cause. Across the dataset, two distinct mechanisms were responsible — and they require different fixes.

The Named Individual mechanism occurs when a founder, physician, or practitioner is featured prominently in the domain, title tag, H1, and body copy simultaneously. The site intends to build personal brand equity. The unintended result is that the individual consumes the majority of the page's entity budget, crowding out service and location signals. The most severe case in the study: a single named individual absorbing 81.7% of total page salience — leaving the city the business serves with a score of 0.037.

The Generic Language mechanism is the more surprising finding and the more widespread one. It occurs when standard marketing copy — the vocabulary of virtually every service business website — contains words that entity analysis consistently classifies as PERSON type. Several sites showed this at extreme salience: one business whose homepage opened with "We Are Builders" had the word builders absorbing 64.1% of total page salience. Another had the phrase "our patients" repeated four times across the homepage, accumulating to a 49.0% PERSON salience score through repetition alone.



The generic language mechanism matters most for this study because it is invisible. A named individual on a homepage is a deliberate choice. Generic language like "our people," "our clients," or "a team of experts" is not — it is the default vocabulary of service business marketing, present on thousands of sites with no awareness of its entity classification consequences.

The highest PERSON salience score in the entire dataset came from a functional medicine site whose domain was the physician's name, whose title tag was "Dr. [Name] is your Functional Medicine Specialist in [City]," and whose body copy referenced the physician throughout. Entity analysis assigned 81.7% of total page salience to that single person — leaving the city name with a salience score of 0.037.

The city was present on the page. It simply didn't register.

The study surfaced several additional patterns worth noting separately from the primary PERSON Entity Trap finding.

Category Misclassification Without a PERSON Trap

Several sites had no PERSON entity in their top signals, but their primary category classification was significantly off-target. A general contractor whose homepage led with "Building Relationships" language was classified by entity analysis under /People & Society/Family & Relationships — not /Business & Industrial/Construction. The page's relational copy was legible to humans; to a classifier, it read as content about personal relationships.

A related case: a construction firm whose mission copy emphasized "our people," "putting people first," and "the best people in the market" received no primary category classification at all. The page was, in the language of entity analysis, topically unresolvable. It wasn't about construction. It wasn't about any recognizable category. It had achieved a kind of perfect invisibility.

Dynamic Title Tag Contamination

One integrative health practice had a rotating blog headline as its page title — the kind of CMS feature that feels like a smart way to keep the title tag fresh. On the day of analysis, the title read: "YOUR CHRONIC BACK PAIN BEGAN WITH YOUR UNHEALTHY GUT." Entity analysis classified the page primarily under /Health/Pain Management at 0.62 confidence — not /Health, not /Medical Facilities & Services. The practice was a general integrative medicine clinic. Its homepage was being understood as pain management content.

Testimonial Copy as an Uncontrolled PERSON Signal

One multi-specialty practice had embedded Google review copy on its homepage — a standard social proof implementation. One review mentioned a physician by name. Entity analysis picked up that name, classified it as a PERSON, and assigned it the highest salience score on the page at 17.2% — above the practice name, above the service descriptions, above the city. A testimonial about a positive patient experience was the most entity-salient element on the page.

Across the full dataset, the following language patterns were confirmed to fire PERSON entity classifications at meaningful salience levels. Many of these are single words that appear naturally in any service business's homepage copy.

Generic collective nouns that fire as PERSON: people, patients, clients, builders, doctors, experts, employees, clinicians, healers, physicians, members, technicians, practitioners, partners

Structural patterns: "We Are [noun]" in headlines, "I am [value list]" in hero copy, "You are [descriptor]" in mission statements

Compound patterns that amplify salience: Repetition of any of the above across multiple sections of the same page (confirmed to accumulate salience additively — one practice with "our patients" appearing four times across the homepage had patients as its top-ranked entity at 49.0% salience)

Structural amplifiers: Any of the above appearing in a title tag, H1, or domain name will dramatically increase the salience assigned relative to the same word appearing only in body copy.

Entity analysis of your own homepage will reveal whether any of these patterns are active. At minimum, look at what entity is ranked #1 on your page — and what type it has been assigned.

If the top entity is classified as PERSON and it is not the name of a real individual intentionally featured on the page, the PERSON Entity Trap has likely fired.

If no location entity appears in your top five results, and your business depends on local search visibility, Salience Budget Collapse may be suppressing your geographic signal.

The fix, in most cases, is not a complete rewrite. It is a targeted revision of the specific phrase driving the misclassification — often a single sentence in the hero section or the headline — combined with an inversion that leads with the location or organization name before introducing service language.

These findings matter beyond the specific businesses that were analyzed. They describe a structural tension between how service businesses write about themselves and how entity classifiers interpret that writing.

Service businesses — especially those in healthcare, trades, and professional services — are trained by marketing convention to lead with people. "Our team of experts." "The practitioners who listen." "Builders who care about craft." This language is human, warm, and relatable. It converts well. It reads as trustworthy.

To an entity classifier, the most salient thing about these pages is that they are about a person. Or a group of people. Not a service. Not a city. Not an organization.

The language that makes service businesses sound human to their customers may be exactly what causes search infrastructure to misclassify what they do and where they do it.

That gap — between human-readable intent and machine-interpretable signal — is not unique to SEO. It is a problem that will compound as entity analysis, and AI-driven search interpretation play larger roles in determining which businesses get found. The businesses that understand it will have a structural advantage. Most of them don't know it exists.

This research analyzed 85 valid small business homepages across the construction, trades, and functional medicine sectors. Entity analysis was performed on homepage copy only. Sites returning no readable body copy due to JavaScript rendering, maintenance pages, or malicious content injection were excluded from the dataset and replaced where possible; 14 DQ sites were excluded in total. Analysis was conducted in Q1–Q2 2026.

Joe Provence is the co-founder of Salt Creative.