We ran 309 free audits this week — here's what's broken on the median local business site

We run free audits at arlo.sh. In the trailing seven days, we've completed 309. This is what the data shows.

The median local business site has no schema markup, a GBP that hasn't posted in over two months, and zero AI citations on its own brand and service queries. Half of sites are worse.

Here are three signals, each broken down from our audit corpus.

Signal 1: Schema coverage

Schema markup is how AI engines identify a local business with confidence. Without it, ChatGPT and Perplexity fall back to aggregators — Yelp, Healthgrades, Zocdoc — instead of your site.

Schema distribution across 309 audits (Arlo corpus, May 2026):

No schema at all — 41% (127 sites). The engine is guessing from prose and third-party mentions only.
Generic LocalBusiness only — 28% (87 sites). Entity recognition exists, but no service-level or Q&A extraction. The AI knows the business exists; it doesn't know what it does.
LocalBusiness + FAQPage — 18% (56 sites). Partial Q&A extraction begins here. Answerable queries start showing up in citations.
Full stack (LocalBusiness + FAQPage + Service + Article) — 13% (39 sites). The only tier with consistent AI citation confidence. Median schema score at this tier: 81/100.

Median schema score across all 309 sites: 29 / 100.

Only 1 in 8 sites has enough schema for an AI engine to cite it with confidence. The other 7 are competing on prose alone — against aggregators that have had structured data for a decade.

Signal 2: GBP posting cadence

GBP recency is the largest local-pack signal after entity-to-searcher distance. A profile that posts weekly outranks a profile that went dormant in February — regardless of review count or years in business.

GBP cadence distribution across 309 audits:

No GBP or unclaimed — 22% (68 sites). No local-pack presence at all. AI engines have no GBP signal to pull from.
Last post > 90 days ago — 35% (108 sites). Actively sending an inactivity signal. Google treats a 90-day gap as a soft "may have closed" indicator.
Last post 30–90 days ago — 19% (59 sites). In the recency-decay window. Not penalized, but earning no cadence points.
Last post 8–30 days ago — 14% (43 sites). Acceptable. Compounding hasn't started yet.
Weekly cadence (≤ 7 days) — 10% (31 sites). Actively compounding. This tier averages 2.8x the profile views of the 90-day-dormant tier in our corpus.

Median days since last GBP post: 68. The typical site is one month inside the penalty-risk window.

The 10% posting weekly are building a durable advantage. Our managed clients confirm the trajectory: Vista Chiropractic reached 3.2x GBP views in 90 days from a weekly cadence restart. Bright Smile Dentistry reached 47 ChatGPT citations in four months, starting from a dormant profile. Both started at the median.

Signal 3: AI citation rate

We test five engines per audit — ChatGPT (GPT-4o), Perplexity sonar-pro, Gemini 1.5, Claude Sonnet, and Google AI Overviews — against five brand and primary-service queries per site. Maximum score: 25 citations.

Citation distribution across 309 audits:

0 / 25 — 74% of sites (229). Not cited once, on any engine, on any query. Completely invisible to AI search.
1–5 / 25 — 15% (46 sites). At least one citation exists. Usually brand-name queries only, not service queries.
6–15 / 25 — 8% (25 sites). Multi-engine coverage emerging. These sites almost always have schema and an active GBP.
16–25 / 25 — 3% (9 sites). Full AI-search presence. Every site in this tier has a complete schema stack, weekly GBP cadence, and answer-first content published within the last 30 days.

Median citation score: 0 / 25.

No site in the corpus with zero schema markup reached 6+ citations. Schema is a prerequisite, not a booster — AI engines need structured entity data before prose quality even enters the picture.

The pattern at the top

The 9 sites sitting at 16–25 citations share three observable properties:

Full schema stack — LocalBusiness specialized to vertical type (Dentist, Chiropractor, MedSpa), FAQPage on every service page, Service entities, Article on all content.
GBP posted within 7 days, maintained continuously for 90+ days.
Two or more answer-first blog articles per week, published within the last 30 days.

No single property alone is sufficient. All three together are necessary and, based on this corpus, appear to be sufficient for reaching the top citation tier.

What comes next

We'll publish this breakdown every quarter — same corpus methodology, same five signals — so you can benchmark your position as AI search matures through 2026.

If you want to see where your site lands against these 309, run a free audit at arlo.sh.

*Data source: Arlo audit corpus, trailing 7-day window ending May 4, 2026. N=309. Schema detection: automated structured-data parser. GBP cadence: Google Places API last-post timestamp. AI citation checks: live queries to ChatGPT-4o, Perplexity sonar-pro (return_citations=true), Gemini 1.5 Flash, Claude 3.5 Sonnet, and Google AI Overviews API.*