TL;DR
- Search Engine Land reported on 8 September 2026 that a recovered Google binary exposed 72 named ranking signals from an internal Maps system called Oyster Rank.
- 25 of the 72 signals are explicitly marked deprecated, alongside 793 data source providers, 446 local search intent types and 10,936 searchable Geostore declarations.
- Named signals include Google reviews, entity query volume, listing impressions and opens, direction requests, website clicks, chain membership, Wikipedia signals, landmarks and road usage.
- The column states the team recovered signal names, not weights, so no signal on the list is proven to carry any particular influence.
- Paris testing of 5,083 API calls and 86,584 results found a median search distance of 6.87 km with geography applied and over 4,000 km without it.
Search Engine Land published a technical column on 8 September 2026 reporting that a research team recovered a Google binary that exposed a non-public scope of Geostore, the database Google uses to represent real world places, and pulled 72 named ranking signals out of an internal Maps ranking system called Oyster Rank. Twenty-five of those 72 signals are explicitly marked deprecated. The signal names are now public knowledge. The weights behind them are not, and the column says so in plain terms rather than leaving the reader to assume otherwise.
That distinction is the whole story. Local ranking advice is usually assembled from correlation studies, agency case notes and Google's own carefully vague help pages. A named list of fields from inside the system is a different class of evidence. It proves a signal exists in the vocabulary. It proves nothing about how much that signal moves a result.
What the researchers actually recovered
According to Search Engine Land's 8 September 2026 column, written by Olivier de Segonzac, the team obtained a binary exposing Geostore and then cross-referenced it against Maps protocols, network traffic, the web index, mobile services, style tables, on-device components and Google's 2024 documentation leak. The cross-reference matters because it lets a field recovered from the binary be read alongside a description that already existed publicly: if a field appeared in the 2024 documentation, the associated description can be read next to it.
The recovered material was not limited to the ranking signals. The column reports 793 data source providers, 446 local search intent types and 10,936 searchable Geostore declarations alongside the 72 Oyster Rank signals. Those numbers describe the size of the surface, not its behaviour, but they are useful context: the places database is a much larger machine than the handful of fields visible in a Google Business Profile.
The named signals
The signal names reported in the column cover things local SEO practitioners already talk about and a few they do not. Google reviews appear as a named signal. So do web query volume for the entity, listing impressions, listing opens, direction requests and website clicks. Beyond behaviour, the list includes chain membership, Wikipedia signals, popularity and prominence measures, landmark information and road usage data.
Read as a group, these fall into three rough families. There are engagement measurements taken inside Google's own surfaces, such as impressions, opens, direction requests and clicks through to a website. There are demand measurements taken outside them, such as query volume for the entity. And there are entity attributes that describe what a place is and where it sits in the world, such as chain membership, Wikipedia presence, landmarks and roads.
The column is careful about what a name proves. It states that recovering the field for Google reviews shows that reviews belong to the Oyster Rank vocabulary, and that this does not prove reviews currently carry a particular weight in a Maps search. The same caution applies to every other name on the list. As the column puts it, the important limitation is that the team recovered the signal names, not their current weights, and the coefficients that would say how much each signal contributes sit outside the scope that was recovered.
A quarter of the list is marked deprecated
Twenty-five of the 72 signals carry an explicit deprecation marker. That is the detail most likely to get lost in the retelling, because a list of 72 signals is a more shareable headline than a list of 47 live ones and 25 dead ones.
The practical consequence is that seeing a name in the recovered set is not, on its own, a reason to build a tactic around it. A signal can exist in the schema, be visible to anyone reading the binary, and still be switched off. Anyone quoting a specific signal from this research as a ranking factor needs to say which side of the deprecation line it falls on, and the burden of that check sits with whoever is making the claim.
The distance test, and why the two numbers are not what they look like
The column adds original testing rather than only reporting the recovered fields. The author ran 5,083 API calls producing 86,584 results in Paris, on queries including the French word for pharmacy and the supermarket brand Carrefour, to measure how heavily geography is weighted in the results Google returns.
With geographic context applied, the median search distance was 6.87 km. With geography removed, it jumped to more than 4,000 km. A dense query such as pharmacy produced a far smaller search area than a brand query such as Carrefour, and testing in a sparsely populated rural area saw the pharmacy query expand significantly. That is the intuitive part: in a city with a pharmacy on most blocks, Google does not need to look far.
The counter-intuitive part is what happened to the ordering. The column reports that when geographic weighting was removed, the non-geographic order remained extremely stable. In other words, proximity was governing how wide a net Google cast, not who won inside the net. The author's conclusion is that distance remains fundamental in local SEO, but that the model of "I am closer, so I should rank higher" is incomplete.
Two caveats have to travel with these numbers. All of this testing was run in Paris on French-language queries. The figures are a measurement of one city's density and one language's query behaviour, and they must not be presented as measured in Thailand or anywhere else. The signal list itself carries no such limit, because a field in Geostore is a field in Geostore regardless of which market a search comes from.
What is recovered fact and what is the author's reading
The column moves from recovered fields to practical conclusions, and the two need separating. The recovered facts are the counts and the names: 72 signals, 25 deprecated, 793 data source providers, 446 local search intent types, 10,936 searchable declarations, plus the field names themselves. The distance figures are the author's own measurements, described with their method attached.
The advice that follows is architectural interpretation. The column argues for building a complete entity representation that goes beyond the Google Business Profile, for semantic completeness, for consistent brand and location information across a business's web properties, and for making it easy for Google to connect relevant attributes and evidence to the business. It raises the questions of whether Google sees real demand for the brand and what reviews consistently say about specific aspects of it. Those are reasoned positions built on the recovered architecture. They are not statements Google has made, and they are not weights anyone has measured.
This table sets out the counts the column reported, which are the parts of the story that can be quoted without interpretation.
| What was recovered | Count reported on 8 September 2026 |
|---|---|
| Named Oyster Rank ranking signals | 72 |
| Of those, explicitly marked deprecated | 25 |
| Data source providers | 793 |
| Local search intent types | 446 |
| Searchable Geostore declarations | 10,936 |
What the column did not say
Google has not commented on the research in the material reported. There is no statement from Google confirming that Oyster Rank is the current production system, no publication of weights, and no timeline attached to the deprecated signals. The column does not claim that any of the 72 signals is more important than another, because the coefficients were outside what the team recovered. Nothing in it addresses Thailand, Southeast Asia or any non-French market, and no testing outside Paris is reported.
It also does not turn the deprecation count into a list of tactics to stop doing. That translation step, if anyone makes it, is theirs and not the source's.
What this means for marketers in Thailand
The most useful takeaway is a reframing rather than a task list. If proximity mostly sets the radius and the ordering inside that radius is driven by other signals, then a business competing in a dense Bangkok district is competing on the non-geographic part of the system against everyone else within the same short radius. Adding branches changes the radius question. It does not answer the ordering question.
The named signals suggest where the ordering evidence comes from, with the weighting caveat attached throughout. Engagement inside Google's surfaces is measured: impressions, opens, direction requests and clicks to the website all appear as named fields. Demand outside Google's surfaces is measured too, in the form of query volume for the entity. That is a brand-building argument reaching a channel that is usually treated as purely technical, and it is worth taking seriously precisely because it is not new advice dressed up as a leak. For a business auditing its own local SEO position, the honest version of the checklist is: is the profile complete and specific, does the website describe the same entity the profile describes, and is there any measurable search demand for the brand name at all.
The Wikipedia, chain membership and landmark signals are a reminder that Google is assembling a picture of a place from sources a business does not control. That is not something to game. It is a reason to make the sources a business does control agree with each other, which is the same discipline that underpins any serious SEO programme in Thailand: one canonical set of facts about the business, repeated consistently everywhere those facts appear.
What should not happen is a wave of Thai local SEO advice quoting 6.87 km as though it were a national constant. It is a median measured in Paris. Bangkok density, Chiang Mai density and rural provincial density are all different problems, and none of them were measured here.
What to check in your own listings this week
- Whether your Google Business Profile categories, attributes and service descriptions actually describe what the business does, rather than repeating a keyword.
- Whether the website's own pages corroborate the profile: same name, same address format, same services, same locations.
- Whether anyone searches for the brand by name, which is measurable in Search Console and in Google Business Profile insights.
- What reviews say repeatedly about specific aspects of the business, not just how many stars they carry.
- Whether multi-location listings are consistent with each other, since chain membership appears as a named signal.
Frequently asked questions
Did Google confirm these 72 Maps ranking signals?
No. The signals were extracted from a recovered binary by an outside research team and reported by Search Engine Land on 8 September 2026, not published or confirmed by Google. No Google statement on the research appears in the reporting.
Do the 6.87 km and 4,000 km distance figures apply in Thailand?
No, they were measured in Paris on French-language queries and cannot be read as Thai numbers. The column reports 5,083 API calls and 86,584 results from Paris testing, plus a sparsely populated rural comparison; it does not report any testing in Thailand or anywhere in Asia.
Which of the 72 signals are the deprecated ones?
The column states that 25 of the 72 carry an explicit deprecation marker, and that count is the fact to quote. Anyone building a tactic on a specific named signal should establish whether that signal is one of the live ones before treating it as actionable.
Does this tell me how much reviews matter for Maps rankings?
No. The column is explicit that recovering the Google reviews field proves reviews belong to the Oyster Rank vocabulary and does not prove reviews currently carry a particular weight in a Maps search. Weights were outside the scope the team recovered.
Do I need to change anything in my Google Business Profile because of this?
Nothing in the research obliges an immediate change, because no weights were recovered and Google announced no product change. The recommendations in the column, around entity completeness and consistency between a profile and a website, are the author's interpretation of the recovered architecture rather than instructions from Google.
The short version
A named list beats a vibe, and this is a named list. It is also a list where a quarter of the entries are marked dead, none of the entries carry a weight, and the most quotable numbers in the write-up were measured in a single European city. Read it as a map of what the system can see, published by Search Engine Land on 8 September 2026, rather than as a set of dials anyone can now turn.
If you want a second opinion on how your locations are actually represented across Google's surfaces, and where the gaps between your profile, your site and your brand demand sit, Relevant Audience runs that audit as part of its local search work.
