Entity Trust: does the machine know who you are?

Two words carry half of that statement. “Can help” is Google's own wording and no hedge of ours: the page on organization markup names two purposes — understanding the administrative details and telling the organization apart from other organizations — and puts a modal verb in front of both. The knowledge panel help, in turn, describes what such a box is meant for: a quick overview of a topic, based on what Google understands from the content available on the web. That, conversely, every entry in the Knowledge Graph leads to a display is not something it says.

Neither document says anything anywhere about grades, point scores or trust numbers. Neither the knowledge panel help nor Google's page on organization markup comments on a rating at all. Anyone who reads “entity trust” as a number is looking in these two sources for something that is not there.

The document that comes closest to the rating question is a third one from Google, and what it says is uncomfortable. E-E-A-T itself is not a specific ranking factor, Google states; what is useful is a mix of factors by which content with good E-E-A-T can be identified. For topics that carry particular weight the same page adds that Google's systems give even more weight to content that aligns with strong E-E-A-T. “Not a ranking factor” therefore does not mean “without significance” here — it means there is no dial by that name.

The same page also places the Search Quality Rater Guidelines, the criteria catalogue used by trained human raters. Those raters, Google writes, have no control over how pages rank, and their data is not used directly in the ranking algorithms. The word “directly” stands like that in the original and belongs to the sentence: it does not say the data is not used at all.

For the question of where the same identity appears outside your own website, the vocabulary holds an entry of its own. Schema.org defines sameAs as the address of a reference web page that unambiguously indicates the identity of an item, and names as examples the item's Wikipedia page, Wikidata entry or official website. What setting this property brings about is not stated on that page — the fourth section returns to this.

The first question is therefore not what grade you receive, but whether an entry about you exists at all. That sharpening is ours and not a quotation; none of the three Google documents named here puts it that way. It is the order in which this article sorts the next four sections.

The graphic is built the same way: four sources on the left, one entity node in the middle, two states of assignment on the right. No point score, no ranking. That contradictory details increase uncertainty is a model assumption of this article — where it comes from and where its limit lies is set out in the third section.

Google recommends bundling the details about the organization rather than scattering them: on the home page, or on a single page that describes the organization, for example the about us page. They do not have to appear on every page of the site under that recommendation. What to focus on the same page lists as examples: the name of the organization, plus an indication of real-world presence — address and telephone are named — and one of online presence — website address and logo are named.

1. 01 · Company

   Who publishes the website?

   Schema.org defines Organization as an organization such as a school, an NGO, a corporation or a club, and places the type in the hierarchy directly below Thing. For the official name there is a property of its own: legalName is the official name, for example the registered company name, while name there is simply the name of the item. Anyone who knows both fields can tell a trading name from a register entry without losing either detail.

2. 02 · Place

   Where is the business located, and whom does it serve?

   For the particular physical business or the branch of an organization, Schema.org holds its own type LocalBusiness; the examples range from a restaurant through a branch of a bank to a medical practice. For the Google Business Profile, Google requires a precise, accurate address or a service area and states that post office boxes at remote locations are not acceptable. A separate bullet point says: create the profile for your actual, real-world location. Both apply to the profile and not to the location details on your own website.

3. 03 · Services

   What is actually being offered?

   Schema.org holds Service as its own type for a service provided by an organization, with delivery and print services as examples. In the Google Business Profile the category has to be distinguished from this, and the rule for it sits under the heading “Categories”: the category should complete the sentence “This business IS a” and not “this business HAS a”; it describes the business holistically and not the list of its services, products or amenities. Directly above it stands the addition not to use categories merely as keywords.

4. 04 · People

   Who stands behind it by name?

   Schema.org links a person to an organization through the property affiliation, for instance a school, a club or a team. It is not the only link: the same table also holds worksFor, alumniOf and memberOf. In Google's recommended organization properties, by contrast, “people” does not appear as a named kind of detail of its own. That this article carries them as a fourth kind is our own sorting.

One property from the same table is regularly overrated. knowsAbout names a topic that someone knows about — according to Schema.org primarily a person and “less typically” an organization. The entry suggests possible expertise without implying it, and the vocabulary expressly does not distinguish skill levels. A field with which expertise could be asserted is not what this is.

And in the organization markup, none of these details is a required field. Google writes about organization markup that there are no required properties; instead one should add the properties that apply to one's own organization. Anyone who reads “Google requires” here, where “recommends” is written, builds themselves a rule that does not exist.

For a system to bring these details together at all, they have to arrive on retrieval. Whether name, location and service sit in the delivered document or only appear afterwards is therefore no side issue — How AI really reads your website puts that difference in context.

What this looks like in everyday life is best shown by an example. None of the sources named below describes small businesses in German-speaking countries; the following case is therefore ours and expressly invented. Elektro Kranz in Lichtenberg appears in the commercial register as “Kranz Elektroinstallation GmbH”, the shop sign reads “Elektro Kranz”, and in the business profile the firm is listed as “Elektro Kranz Berlin”. Three spellings, one business. Which detail belongs where is decided by exactly the fields above: legalName for the register entry, name for the name in use — and in the profile the real-world name as customers know it.

At Google it starts with the name. The name in the business profile should reflect the business's real-world name — as used consistently on the storefront, on the website and on the stationery, and as known to customers. For organization markup the same thought stands one step weaker: Google recommends there using the same name and alternate name that the site already uses as its site name. Recommends, not requires — in the same section stands the sentence that there are no required properties.

Google draws a hard line elsewhere. The structured data guidelines say that content which is not visible to readers of the page should not be marked up: if the JSON-LD describes a performer, the HTML body must describe that same performer. The machine-readable detail and the visible text should describe the same thing — that is consistency not between two websites but within a single page.

What inaccurate details can cost, Google states for the business profile itself. Businesses with complete and accurate information are more likely to show up in local search results; if the information is not accurate, the profile might not show up for relevant searches in your own area. Both are probability statements, not commitments. For pointing to your own profiles at third parties, organization markup provides the sameAs field; named as examples are the profile page on a social network and the one on a review site, and several addresses are possible.

And what if a detail is disputed? For the Knowledge Graph, Google expressly reserves the right not to act — on facts that are reasonably disputed or lack demonstrative evidence. The sentence sits in the correction and removal policy of the Knowledge Graph and applies to it. It is not a general statement about what Google does with contradictory business details.

On naming, it is put more directly in Microsoft's help documentation. The Bing Webmaster Guidelines carry a section of their own, “Define Entities Clearly and Consistently”, and it opens with an imperative: use clear and consistent naming for people, organizations, products and locations. The second of exactly two bullet points in that section gives the reason — clear entity definition improves grounding visibility and citation accuracy. By how much it improves is not stated there; the verb is “improves”, and Microsoft names no order of magnitude at this point. In the help on the AI Performance report, consistency across formats is added: text, images and other media should describe the same products, entities and concepts.

The mechanism behind this comes from a blog post, and the type of document belongs to the statement. “Evolving role of the index” is a position piece from Microsoft's AI group with a named byline — not product documentation and not a guideline for website operators. What it describes is kept general: when two indexed sources contradict each other, a grounding index cannot simply surface one above the other and move on; it needs to register that conflict. The reason stands in the same sentence, as a subordinate clause and in the possible mood — an AI system that silently arbitrates between contradictory sources is one that may confidently assert the wrong thing.

In the comparison in the same section, contradiction is a measure of its own, and the two columns say different things. For traditional search it states that one source can be surfaced above another and the user left to arbitrate; for grounding, conflict must be detected and represented, because silent arbitration risks confident wrong answers. Under a different heading the same post describes grounding as determining what information can responsibly support a claim, and as having the discipline to withhold when the evidence is not there. The closing paragraph puts the separation into two sentences: search optimizes for likelihood of relevance, grounding must measure strength of evidence. Abstention is a valid outcome when support is missing, stale or conflicting — a valid outcome, mind you, and not an obligation.

The jump from there to your case is ours. Microsoft speaks of a grounding index and of two indexed sources; the case “your own website contradicts your own business profile” does not appear in Microsoft's post. That such a contradiction makes assigning a business harder is therefore a model assumption of this article and not a vendor statement. It is plausible, and it stands here as what it is.

No commitment follows from this. On the five Google pages this section quotes from — the organization markup, the structured data guidelines, the profile guidelines, the local ranking tips and the Knowledge Graph help — not one sentence links consistency to a position in search. Matching details therefore secure no position. Under the model above they spare a system a decision it would otherwise have to make — and a problem only arises where the same detail reads differently in two places. That Elektro Kranz is named differently in the register than on the shop sign is therefore not one of those cases: these are two different details in two fields provided for them.

The soberest sentence on the subject sits in Google's general structured data guidelines: using structured data enables a feature to be present — it does not guarantee that it will be present. How far the separation of markup and placement goes is shown by the same page in a special case: a structured data manual action costs a page its eligibility as a rich result, but does not change how it ranks in web search.

For the generative features of Google Search it becomes clearer still. Structured data is not required for that, Google's guide writes, and there is no special schema.org markup you would need to add for it. It is nevertheless a good idea to keep using it as part of ordinary search engine work, because it helps with being eligible for rich results on Google Search. “No markup you need to add” is a condition here and not a denial of existence: that no suitable vocabulary exists is not what the sentence says.

Microsoft says something different at this point, and the difference stays where it is. The Bing Webmaster Guidelines carry a section of their own on structured data: it may support clearer grounding but does not guarantee visibility or grounding traffic, and the markup must accurately reflect visible content. In the abuse section of the same guideline stands the reverse side — markup that is irrelevant, inaccurate or misleading may be ignored and can affect trust and eligibility for enhanced features. Google writes “not required”, Microsoft writes “may support clearer grounding”. Both are caveats, but they do not say the same thing, and no shared vendor line follows from them.

Two fields of organization markup lead away from your own website. sameAs takes several addresses and points to pages on other websites with additional information about the organization; at Schema.org it is not a property of the organization at all but one of the supertype Thing, and on the Organization page it sits in the section “Properties from Thing”. The second field is vatID: Google calls the VAT code an important trust signal for users, if it applies to the country and the business at all, and gives as an example in brackets that users can look the business up in public VAT registries. By what means trust arises there, the page does not say.

Neither Schema.org's page on sameAs nor Google's page on organization markup says what setting this property brings about. How many addresses make sense, which target addresses are evaluated, or whether they are evaluated at all is not stated on either of them. This is the largest open point of this section, and it belongs in the text and not in a footnote. Google states an effect only for organization markup as a whole, and there too in the mode of the possible: it can help to understand the organization better and to tell it apart from others.

The logo belongs in this row as well, and Google frames it as a possibility. Under the heading “Add structured data” it says one can specify which image Google should use as the organization's logo in search results and in one's own knowledge panel; for that one adds organization markup to the official website that identifies the location of the preferred logo. “You can specify” is the wording. No obligation stands there — and that the markup brings about inclusion in the Knowledge Graph is not something the sentence says either.

The second half of the heading leads away from your own website. Three kinds of external reference appear in the documents named below, and they work differently.

• Registers and identifiers

  A number held on file elsewhere

  The VAT code and register identifiers under ISO 6523 point to an entry outside your own website. Schema.org advises preferring ISO-6523 identifiers to vatID or taxID where possible, because ISO identifiers are less ambiguous. Google expressly calls the VAT code an important trust signal for users. Whether such an identifier is evaluated by search engines or AI systems is not something this article evidences.

• Wikidata

  A knowledge base with admission criteria

  Wikidata describes itself as a secondary knowledge base that records not just statements but also their sources and connections to other databases, and thereby supports the notion of verifiability. An item is only acceptable there if it meets at least one of three criteria: a valid sitelink into a Wikimedia project, a clearly identifiable entity that can be described with serious and publicly available references, or a structural need. Data outside these criteria is, by its own account, not accepted; the page points instead to the sister projects Wikibase Cloud and Wikibase Suite. It calls its own criteria intentionally left a bit vague and points to an ongoing reform discussion.

• Wikipedia

  Notability that others decide on

  The English-language notability guideline for organizations says that no company and no organization is inherently notable; no organization is exempt from this requirement. Under the primary criteria a company is presumed notable if it has been the subject of significant coverage in multiple reliable secondary sources that are independent of the subject — a sufficient condition, alongside which the same page names alternative routes. The German-language Wikipedia lists thresholds of its own for commercial companies, among them at least 1000 full-time employees or annual revenue of more than 100 million euros, and writes in a box of its own about them: sufficient, not necessary conditions.

One point about this is regularly underestimated in practice. Wikipedia sums up its conflict-of-interest rule in a box: do not edit in your own interests or in the interests of your external relationships, and disclose paid editing. In the introductory section of the same page stands the reason — such editing is strongly discouraged, it undermines public confidence and risks causing public embarrassment to the individuals and companies being promoted. Advertising is prohibited on Wikipedia under the official policy linked there, and the procedure against it has three steps: neutralize, remove the promotional content, and if necessary nominate for deletion. Only as an exception to this can an article that contains only blatant advertising, with no other useful content, be tagged for speedy deletion.

That an entry at Wikidata or Wikipedia brings about inclusion in Google's Knowledge Graph is not stated in this article. It carries no evidence for that effect, and it has not investigated it either. What is evidenced is something weaker and still usable: these two directories have admission rules, the rules are publicly readable, and they were not set by the company they concern.

Reading an outside mention as proof of authority likewise goes beyond the sources named below. The closest to it is Google's description of prominence in the local context: prominent places are more likely to show up in search results, and this factor is also based on information such as how many websites link to the business and how many reviews it has. That is the description of a factor for local results and not a statement that mentions elsewhere establish authority. Prominence is nowhere set apart from signals on your own website on that page; under “Relevance” the opposite stands, namely the call to provide complete and detailed business information.

This article keeps two things apart that regularly collapse into one another in talk about entity trust. A clear signal is a detail that you set, maintain and can read back at any time. An observed vendor result is what a named system actually showed afterwards. The first can be produced, the second only established. A source that sets exactly these two terms against each other does not exist among the documents named below — the distinction is ours.

• clear signal

  What you set and maintain yourself

  The name in the markup, the address in the business profile, the references in the sameAs field, the logo you name through organization markup. These details stand fixed, they can be read back, and they change only when you change them. Alongside that, Google describes how its algorithms find information such as your site's name, corporate contact information and social profiles that is publicly available on the web — and that anyone verified as an official representative can update the knowledge panel and override the information Google finds automatically.

• observed vendor result

  What a system shows afterwards

  Microsoft shows in the AI Performance report of the Bing Webmaster Tools which content was cited — and states in the same documentation that the report does not measure rankings, authority, performance or importance. The values shown are a sample of overall citation activity and may still be refined as further data is processed. How large the part not captured is is not stated there: the page names neither share nor coverage nor margin of error.

• without a commitment

  What remains open

  Whether a particular system names your business in a particular answer cannot, by the documents named below, be established beforehand, and afterwards only for the answers actually observed. For structured data Google expressly does not promise that features which consume it will show up in search results. Microsoft writes that improving content quality, structure and freshness can help strengthen content but does not guarantee a specific citation outcome. On ChatGPT, Perplexity, Claude and the Gemini app the documents named below say nothing; nothing is therefore stated here about them, neither as a claim nor as a denial.

With the knowledge panel the difference becomes clearest. Knowledge panels are created automatically by the search algorithm, Google writes, when there is enough information available on the open web. And further: over time Google Search automated systems will determine when, where and how a specific query should display a knowledge panel; this happens automatically and is not something Google can or would influence. “When, where and how” stands there — “whether” does not appear in that list. A submission or application route is not described by the two Google pages this paragraph quotes from; they know the automatic creation and the takeover of an already existing panel by a verified representative.

For local results Google becomes fundamental at one point: a better local ranking can neither be requested nor bought, and the details of the algorithm are deliberately kept confidential so that the ranking system stays as fair as possible for everyone. What you can save yourself is stated by the guide to generative AI features at one point with a load-bearing adjective: for Google Search you can ignore tactics such as chunking content, creating unnecessary AI text files or pursuing inauthentic mentions. It is about inauthentic mentions. Against mentions as such the sentence says nothing.

On optimization Microsoft is expressly not silent. The help on the AI Performance report has a section of its own with content recommendations and links to a further guide; Microsoft frames these recommendations as general best practices one may consider. Only two things are refused: the promise of an outcome and the causal inference. Citation Share, Microsoft writes, can help you observe whether your share changes over time after content updates — the metric does not thereby establish a causal link.

From a missing commitment nothing follows in the other direction. That Google and Microsoft make no promise of a mention for their own systems does not mean that a mention is unlikely. It means the question can only be answered by observation — and that anyone who assures you of a mention by a particular AI system is promising something that Google and Microsoft, in the documents named below, do not promise for their own systems themselves. The part you hold in your hands is the first one: clear, matching details about company, place, services and people. What a system makes of them is the second.

Google — organization markup

Google names the purpose of the markup, the bundling on one page, real-world and online presence, the naming recommendation, the sameAs and vatID fields, the sentence about the absence of required properties, and the absent display commitment.

Organization (Organization) structured data Page state 15 April 2026, retrieved on 4 September 2026

Google — structured data

Google requires markup and visible text to describe the same thing, and names the sentence about enabling rather than guaranteeing, and the manual action without any ranking effect.

General structured data guidelines Page dated 10 July 2026, retrieved on 4 September 2026

Google — people-first content

Google states that E-E-A-T itself is not a specific ranking factor, names the greater weight for topics that carry particular weight, and places the Search Quality Rater Guidelines.

Creating helpful, reliable, people-first content Page dated 10 December 2025, retrieved on 4 September 2026

Google — generative AI features

Google states that structured data is not required for generative AI search and that there is no special schema.org markup to add, and points to inauthentic mentions as an ignorable tactic.

Optimizing your website for generative AI features on Google Search Page dated 10 July 2026, retrieved on 4 September 2026

Google — business details

Google describes the logo through organization markup as a possibility, as well as the automatically found information and its override by a verified representative. The page carries the title “Add Business Details to Google” and the heading “Establish your business details with Google”.

Add Business Details to Google Page dated 10 December 2025, retrieved on 4 September 2026

Google Help — knowledge panel

The Google help describes knowledge panels for entities in the Knowledge Graph, the automatic creation, the decision of automated systems about display, and the reservation for reasonably disputed facts.

About knowledge panels · How Google's Knowledge Graph works Both pages without a date, retrieved on 4 September 2026

Google Help — Business Profile

The Google help requires the real-world business name and names the rule for choosing categories, address and real-world location, the consequence of inaccurate information, the factors of prominence, and the sentence that a better local ranking can neither be requested nor bought.

Guidelines for representing your business on Google · Tips to improve your local ranking on Google Both pages without a date, retrieved on 4 September 2026

Microsoft — Bing Blogs

Microsoft describes the conflict mechanism in the grounding index, abstention as a valid outcome, the withholding without evidence, and the separation of the target measures of search and grounding. Corporate blog post with a named byline, not product documentation and not a guideline for website operators.

Evolving role of the index: From ranking pages to supporting answers Published on 6 May 2026, retrieved on 4 September 2026

Microsoft — Webmaster Guidelines

Microsoft calls for the clear and consistent naming of entities in the imperative and cites the sentence about grounding visibility and citation accuracy, the limit of structured data, and the section on misleading markup.

Bing Webmaster Guidelines No date given on the page; the article text only becomes visible through JavaScript. Retrieved on 4 September 2026

Microsoft — Bing Webmaster Tools

The Bing Webmaster Tools help names the consistency across formats, the four self-limitations of the AI Performance report, the sample character of the values, the excluded causal inference, and the framing of the content recommendations as general best practices.

AI Performance in Bing Webmaster Tools No date given on the page; the article text only becomes visible through JavaScript. Retrieved on 4 September 2026

Schema.org — Organization

Schema.org defines the type and names the hierarchy line below Thing, the properties legalName, knowsAbout and iso6523Code, the placement of sameAs under “Properties from Thing”, and the definition of sameAs as an identity reference.

Organization — Schema.org Type · sameAs — Schema.org Property Without a page-specific date, site version V30.0 of 19 March 2026; the Organization page identified itself as a development version when retrieved. Retrieved on 4 September 2026

Schema.org — further types

Schema.org lists LocalBusiness as the type for the physical business or branch, Service as the type for the service provided, and affiliation as one link between person and organization.

LocalBusiness — Schema.org Type · Service — Schema.org Type · Person — Schema.org Type No page-specific date, site version V30.0 of 19 March 2026, retrieved on 4 September 2026

Wikidata

Wikidata names the three admission criteria for items, points to the sister projects for data outside those criteria, and describes itself as a secondary knowledge base. Community guideline and introduction page of an open wiki respectively, not a vendor commitment.

Wikidata:Notability · Wikidata:Introduction Last edited on 30 August and 31 July 2026, retrieved on 4 September 2026

Wikipedia (English)

Wikipedia states that notability is not inherent and names the primary criteria as a sufficient route, the advertising ban together with its three-step procedure, and the conflict-of-interest rule with its reasoning. Both are guidelines with an express exception clause, not policy.

Wikipedia:Notability (organizations and companies) · Wikipedia:Conflict of interest Last edited on 30 and 18 August 2026, retrieved on 4 September 2026

Wikipedia (German)

Wikipedia names the alternative notability criteria for commercial companies and the box that marks them as sufficient, not necessary, conditions. Only the section on commercial companies was consulted.

Wikipedia:Relevanzkriterien Last edited on 15 August 2026, retrieved on 4 September 2026

State of this article: 4 September 2026. Vendor statements can change without notice; a measurement stays bound to its date. Where this article names a date, that is not a formality but part of the statement. Two of the pages named here show their text only after JavaScript has run — anyone who wants to check needs a browser for that.