12,093 Requests, 6,066 Data Retrievals. Here’s What They Mean.

The StreetProof homepage originally showed 12,093 successful machine retrievals in a trailing 24-hour window. A September 24 audit found that the counter included 6,027 HTTP HEAD checks paired with later GET requests. HEAD returns response headers, not the requested data body, so those checks should not have been called retrievals.

The corrected retrieval total is 6,066. A successful retrieval now means a qualifying external operation that returns public StreetProof data. A crawler family still counts only traffic that matched an active, known crawler signature. A client can read thousands of records without identifying itself as Google, Claude, Meta, or any other recognized family.

Of the 6,066 data retrievals, 6,056 were classified as unattributed automation. Nine were attributed REST/API traffic and one was an MCP tool call. No retrieval in the trailing window matched a known crawler signature.

The raw request event count was real. The first retrieval label was wrong. This correction keeps both numbers and states what each one measures.

StreetProof has a supply side and a demand side. The supply side turns eligible physical-world inputs into accepted observations. The demand side records privacy-safe machine retrievals of the resulting public data.

For index production, the useful rate is accepted observations per 1,000 processed frames. Raw frame volume mostly describes workload. It does not describe public value. In this window, StreetProof accepted 573 observations from 504 entities across 56 cities, producing 1.78 accepted observations per 1,000 frames.

The immediately preceding 24-hour window produced 556 accepted observations at 1.75 per 1,000 frames. Accepted output rose 3.1% and the normalized rate rose 1.4%. Operationally, that is flat—not a breakthrough and not a collapse.

Across the seven completed UTC days shown below, StreetProof accepted 3,025 observations at a weighted 1.47 per 1,000 frames. The daily rate moved between 1.07 and 2.02. That range is the honest baseline for this early period.

The 6,066 retrievals were not spread evenly across the site. Entity records returned data 2,987 times and evidence endpoints returned data 2,919 times. Together, those two route families accounted for 97.4% of successful machine retrievals.

The requests touched 2,977 entity IDs across 135 city slugs. They appeared through 55 privacy-safe client instances. A client instance is an aggregate signature, not a person, company, or guaranteed physical device. Daily one-way hashing and minimized client data prevent permanent visitor tracking.

The eight-day chart shows why one headline number is not a trend. Traffic arrived in bursts. September 17 had one large unattributed wave. September 22 had a larger one. Known crawler and MCP activity followed different patterns.

That shape matters. A smooth climb could suggest broadening demand. Two sharp spikes are more consistent with enumeration: software walking through a large part of the index. Useful? Yes. Proof of adoption? No.

The defensible conclusion is narrow: StreetProof's public index is machine-readable, and an unidentified automated client or clients retrieved a substantial share of it. The request pattern reached thousands of entities and both record and evidence routes. The system handled the sweep without recorded request failures in the frozen window.

The data does not establish that an AI platform used StreetProof in an answer, that search visibility improved, or that index growth caused the traffic spike. It does not identify the requester. It does not turn 55 aggregate client instances into 55 organizations or people.

The same discipline applies to the production side. A higher frame count is not success. More retained candidates are not success. The public output is accepted observations, and the comparable operating rate is accepted observations per 1,000 frames.

This is why the report presents the two systems beside each other without drawing a causal arrow between them. One measures evidence production. The other measures retrieval demand. Their relationship may become meaningful over time, but one unusual day cannot settle it.