Exclusive finding · Findings · 2026-09-16 · By VSNARY | Emmanuel Orta

Google's Maps provider list: 793 names, no Yelp

A recovered Geostore enumeration names every source a fact about a place can come from. The consumer directories that citation services sell by the dozen are absent. The business's own website is present, with the same standing as a purchased feed.

In August 2026 a research team recovered a Google binary carrying a non-public scope of Geostore, the repository that models every place Google Maps knows about. Among the enumerations they published in full is DataSourceProvider.Provider: the list of every source a fact about a place can come from. It has 793 entries. We extracted all of them and searched the list for the directories the local-SEO industry has spent fifteen years telling businesses to be listed on.

Yelp is not in it. Neither is Foursquare, Yext, Bing Places, Apple Maps, Facebook, the BBB, TripAdvisor, Angi, Thumbtack, Houzz or Nextdoor. Zero occurrences of any of those strings across 793 provider names.

What the list is, and what it is not #

The enumeration is internal vocabulary: it names the sources whose observations can be attached to a field of a place record, each with its own provenance and trust level. It is not a list of everything Google ever reads, and an entry being present does not prove it is populated or weighted today. Seventy-two of the 793 are explicitly marked deprecated. The researchers say this plainly, and so should anyone quoting it: names were recovered, weights were not.

Read within those limits, the shape of the list is still informative. The largest group is Google itself: 266 of the 793 names start with GOOGLE, pipelines like LOCAL_WEB_EXTRACTION, LOCAL_ALGORITHMIC_IDENTITY and SA_FROM_MERCHANT_POSTS. The second largest is the state: 129 government, postal, cadastral, survey and municipal sources, from the US Postal Service to the Boulder, Colorado city data feed. Forty are electric-vehicle charging networks. What is left after mapping vendors and tourism boards is a short list of commercial business-data feeds.

InfoUSA and Infogroup are the same company under two of its names; it trades today as Data Axle. Acxiom is another US business-data supplier named, and the full list carries other US data companies too; the rest of the row are directories in other countries and a generic feed format. The list also keeps labels for long-retired sources, so a supplier's name on it does not show Google uses that source today. Google's own business-listings notice names no United States supplier; the evidence is in where Google and AI assistants get business data.

The owner and the website are providers too #

Two of the Google-internal names matter more for a business than any directory. PARTNER_FRONT_END is the merchant's own listing, the Business Profile a business edits itself. GOOGLE_LOCAL_WEB_EXTRACTION is Google reading the business's own website. The site is not a citation to be counted; it is a source in its own right, with the same standing in the schema as a purchased feed.

The same recovered schema shows how sources are reconciled when they disagree. Provenance is kept per field, so a name, a phone and a category can each come from a different provider. Each provider carries a trust level, six steps from UNKNOWN at 0 through YP_FEEDS at 40, the tier for yellow-pages style purchased data, to SUPER_TRUSTED at 64. And the default merge rule is CONFLATION_PICK_FIRST_VALUE: when observations conflict, the first admissible value in priority order wins. Sources do not vote. A stale feed with priority can outrank an edit the owner made yesterday, which is one plausible reason a listing reverts to an old phone number.

What this does to the citation pitch #

The standard local-SEO product is a syndication service: one form, seventy-plus directories, monthly fee, and a report counting how many listings agree. The pitch rests on an assumption that the directory count is an input to the place record. The recovered enumeration does not support that assumption. The consumer directories that syndication services push to are not named as data sources at all.

This does not say directories are worthless. A Yelp page is a page on the web, and Google reads the web; a consistent name and phone across the sites a customer might actually visit is worth having for the customer's sake. It says something narrower and more useful: the count is not the mechanism. Whatever a directory contributes, it contributes as a web page among web pages, not as a provider feeding the record. Paying per directory to move a number that is not an input is paying for the report.

Where a business can move the record from the outside, on this reading, is short: its own listing, its own website, and the government records for its address. The website is the piece a business controls entirely. Check that it states one name, phone and address the same way in the visible page and the structured data, that every sameAs profile it lists links back, and that any parent brand or chain relation is declared from both sides, as in declaring the chain from both sides. Entity SEO is the name for making that source say one unambiguous thing.

One more name on the list, and what it implies #

GOOGLE_SA_FROM_MERCHANT_POSTS reads as service area inferred from merchant posts: the updates a business publishes to its own profile can feed the area it is understood to serve. That is a source a business runs for free and most never touch. It sits in the same enumeration as the national postal services.

Where this can be checked #

The researchers published the full archive, 10,936 declarations including this enumeration, as a searchable index and a downloadable dump; every name in this post is in it verbatim. The archive is at think.resoneo.com/google-map-dissected/archive. The absence claims are string searches over the recovered names; a name spelled in a way we did not search for would be missed, which is why the searched list is printed above rather than summarised.

One instance we measured ourselves shows why the website matters as a source: a newsroom's own site published its longitude without the minus sign, which places the building in China. The website was the source, and the website was wrong.

What this post does not claim #

It does not claim Google ignores Yelp reviews, or that a directory listing has no value for a customer, or that any signal weight is known; the researchers recovered names and stated that the fifteen fields holding the weights were stripped from the scope they had. It does not claim the enumeration is current in every entry: seventy-two entries are marked deprecated and the rest are merely not marked. What it claims is what the list shows. The directories the industry sells by the dozen are not on it, the business's own website is, and the merge rule between them is priority, not consensus.

Questions this post answers

Does this mean Google does not read Yelp?

No. The enumeration names sources whose observations attach to fields of a place record. A Yelp page is still a web page Google reads. What the list shows is that consumer directories are not named as providers feeding the record, while the business's own website is.

Where do the 793 names come from?

From a Google binary recovered by the RESONEO research team in August 2026, cross-checked against Maps protocols and published as a searchable archive with a full dump. The names are verbatim; the weights behind them were not recovered.

Which US business-data feeds are named?

InfoUSA and Infogroup, two names for the company that trades today as Data Axle, plus Acxiom and a few other US data companies. The list also keeps long-retired sources, so a name on it does not prove Google uses that feed today. No US consumer directory such as Yelp appears under any spelling we searched.

What should a local business do with this?

Treat the website as a source, not a citation: one declared name, phone, address and parent relation, stated the same way in the visible page and the structured data, with every profile link pointing back. Keep the owner-managed listing accurate. The rest of the syndication list is worth exactly what a customer reading it is worth.

What does the coordinate example show?

A newsroom's own site published its longitude without the minus sign, which places the building in China. The site was the source of the wrong value, which is why the website deserves more care than any directory count.

Related findings

Comments

Comments are read before they appear. Nothing is published automatically, and no account is needed.

Writing about this? Facts, live figures and marks — every number on that page is dated and traceable to a scan.

All findings · The dataset · How the dataset works