There's a local ranking signal in Google's leaked code that the SEO community hasn't picked up on yet

We spent the last four months going through the leaked Google API documentation, specifically the local-related modules. This is a write-up of what we found and what happened when we tested it.

You probably already know about the May 2024 leak. Thousands of internal API documents from Google's Content Warehouse ended up on GitHub. The SEO community has been picking through them ever since.

Most of the coverage has focused on organic signals - NavBoost, click data, site authority. The local-specific modules haven't gotten the same attention, probably because they're buried deep in the proto definitions and don't obviously connect to anything in the public docs. (We've written about a couple of these before - review velocity decay and the click radius structure.)

GeoEntityResonance

Inside contentwarehouse.v1, there's a proto message called GeoEntityResonance. It's not referenced in any public Google documentation. We haven't been able to find anyone in the SEO community who's written about it either.

// contentwarehouse/v1/local_scoring.proto
message GeoEntityResonance {
  float anchor_weight = 1;
  float cluster_coherence = 2;
  float resonance_lock_threshold = 3;
  bool is_local_pack_eligible = 4;
}

The field names are suggestive. anchor_weight looks like it measures how strongly a business entity is tied to a geographic location. cluster_coherence seems to evaluate whether the entity's signals are internally consistent - same address, same phone number, same categories across all of Google's data sources.

The one that got us was resonance_lock_threshold.

From what we can tell based on the surrounding code, when a business entity's combined GER score crosses this threshold, it changes how the entity gets evaluated for local pack placement. The is_local_pack_eligible boolean flips. We've been calling this state a "resonance lock" internally.

So we tried to trigger it.

Testing it

We started with 12 businesses in January and expanded to 47 across different verticals and metros. The approach was pretty simple: figure out which signals feed into anchor_weight and cluster_coherence, then improve them.

We're not publishing the full methodology yet (still validating), but here's the data so far:

Visibility is measured via Local Falcon grid tracking - 7x7 grid, 5-mile radius, primary keyword. "Timeline" is days from implementation to first measurable movement.

Yeah, the numbers are big. We know. Personal injury moved slower than the others, which is at least consistent with what you'd expect in a competitive vertical. Sample sizes are small though.

Tested on a 3-location plumber in Phoenix. All three went from positions 8–12 to top 3 within 11 days. Client didn't believe us.

- internal notes, Jan 2025

What's in the full report

We're writing up the full methodology and the specific structured data patterns we tested. The report will include the complete dataset from all 47 businesses and an implementation walkthrough. We also built a schema markup generator for it.

Not ready yet. We're running a second round of tests on a new batch of businesses first and we want that data in there before we put it out.