Decentralised News · Agentic Finance · Decision research
By Heath Muchena · Source review 2026-11-14 · Instrument version 1.0.0
What Matters
A supplier should be able to name the system, documented requirement and observed problem behind a proposed fix. Google’s current guidance rejects several supposed AI-search prerequisites. That does not make machine-readable interfaces useless: it means their business case must identify the service or user they actually help. DN’s claim checker tests that boundary without producing a readiness score or ranking promise.
A crypto publisher can buy a useful integration for the wrong reason. A machine-readable file may help a particular agent consume a dataset while contributing nothing to the Google requirement a sales proposal claims it satisfies. The remedy is to separate the target system from the promised outcome before approving the work.
This article audits five claims against current Google documentation. It focuses on procurement of a specific proposed change, not a complete publishing architecture. The original DN instrument records whether a statement concerns Google search eligibility, optional agent interoperability or an unsupported outcome guarantee.
DN Evidence Block
Sources reviewed: 2026-11-14 · Evidence type: Current official Google Search and Search Console documentation, with a bounded browser-agent reference and an original DN supplier-claim audit.
Scope: Google Search generative AI features. Requirements for another crawler, agent, API client or licensed data customer must be checked separately.
- Google’s current optimization guide identifies five broad myth categories. This article keeps that count rather than forcing the six-item claim in the supplied outline.
- Google distinguishes search eligibility from optional agent-facing experiences. Its current Search Console guidance also describes an inclusion control for generative AI features.
- The DN checker evaluates the scope of a proposed claim using the reviewed documentation. It does not determine whether a commercial proposal has been delivered successfully.
Not established by this work: No DN ranking, traffic, Search Console setting or vendor service was tested in this article. The instrument does not inspect a website, predict citations or certify search eligibility.
Original DN instrument · version 1.0.0
DN Search Requirement Claim Checker
Classify one proposed optimization claim by target system, asserted necessity and reviewed documentation. It records a procurement question; it does not score a website.
Enter your assumptions or load a labelled example. Required blank numbers stay missing; unknown evidence remains unknown. Use nonpersonal labels only. This instrument makes no network requests and uses no persistent browser storage.
Use your own inputs
No current result. Complete inputs or load an illustrative example.
A dated interpretation of documented claim scope. It does not fetch pages, inspect Search Console, validate contracts or approve suppliers. Requirements for a named external client need their own primary documentation.
1. A special AI file is not a Google Search prerequisite
Google’s current guide says llms.txt and other special AI files are not required for Google Search visibility and are ignored for that ranking purpose. A file can still have a separate use for a service that actually consumes it.
Ask the supplier to identify the consumer. Is it Google Search, a named research agent, a licensed customer or an internal retrieval system? Then ask for the actual requirement or integration evidence. A claim that every publisher must deploy the same file to appear in AI answers is too broad to approve as a technical necessity.
A useful statement of work might specify a documented data endpoint, a stable identifier policy and a client that successfully reads it. The acceptance test would be the client’s permitted retrieval and correct interpretation. That is an intelligible integration project. Calling the same work a guaranteed Google ranking fix changes the claim and requires evidence the integration alone cannot supply.
2. Mandatory tiny chunks are not a documented ranking rule
The official guide rejects a requirement to split content into tiny pieces for Google’s generative AI features. It does not prescribe an ideal page length. That leaves publishers free to choose structure around the reader’s task.
For decision research, a caveat can be part of the answer rather than optional context. Splitting a rule from its date, region or eligibility condition may make a page easier to fragment while making the extracted claim less reliable. The proposed change should be assessed against that practical risk.
Request an example showing the current reader problem and the proposed improvement. Can a reader find the answer, understand the applicable conditions and follow the source? If a shorter section improves that experience, use it. If the only rationale is an undocumented universal chunk size, require a better basis before buying the rewrite.
3. Rewriting only for AI phrasing is not required
Google says publishers need not rewrite in a special style for generative AI search or capture every wording variation. A content brief should therefore explain the information gain for the audience, rather than merely listing additional query permutations.
For DN, the valuable editorial unit is a distinct decision: which input is missing, what evidence resolves it and what action changes when the answer changes. A second article that rearranges the same facts around a slightly different query can create more maintenance without adding another decision.
Ask for an overlap map before approving a large rewrite or expansion. It should identify the existing canonical article, the genuinely missing question and the original contribution of each proposed page. That map is also useful for internal links and corrections. It is not a mechanism for manufacturing dozens of near-identical pages around a model’s suggested queries.
4. Manufactured mentions are not a substitute for evidence
Google’s guidance cautions against inauthentic mentions, while its spam policies describe practices that can violate Search rules. A promised volume of mentions does not establish that the mentions are authentic, useful or relevant to a reader.
For a research publisher, a more reviewable deliverable is an identifiable contribution: a reproducible calculator, a clearly scoped dataset, an evidence register or a correction that others can inspect. Those assets may be cited because they answer a question; no supplier can responsibly guarantee that particular systems will cite them.
Separate distribution work from evidence work in the proposal. Distribution can have concrete tasks and observed outcomes. Evidence work can have source and reproducibility checks. Do not let a dashboard of placements stand in for either the accuracy of the underlying claim or a documented causal explanation of search performance.
5. Special AI schema is not an eligibility shortcut
Google says there is no special schema.org markup required for generative AI search. Its structured-data policies require markup to represent the page faithfully for the relevant supported features.
A schema project still needs an explicit target. If the goal is eligibility for a documented rich-result type, use that type’s actual requirements and verify that the visible page supports the claims. A source list should not become an empirical Dataset merely because the word sounds useful, and an unperformed test should not become a review rating.
For an agent integration, define what fields the client needs and how it will handle unknown, expired or conflicting values. That can justify structured data independently of Google. The acceptance test is correct consumption of truthful information. Passing a syntax validator does not prove factual accuracy, ranking improvement or buyer trust.
Methodology, limitations and corrections
DN read the official sources on 14 November 2026. The five sections correspond to the five myth categories in the current Google guide. The checker adds an original procurement workflow: identify the target, distinguish eligibility from an optional integration, and require a reviewable acceptance test.
There are real eligibility topics. The AI-features guidance requires a page to be indexed and eligible to appear with a snippet. The current Search generative AI control documentation describes an inclusion setting and inheritance across properties. This article does not inspect or change DN’s settings.
| Proposed work | Useful acceptance evidence | Claim it cannot establish alone |
|---|---|---|
| Fix an indexing blocker | Identified blocker and verified resulting site state | Guaranteed ranking or citation |
| Publish a client-readable dataset | Named client consumes correct, dated fields | A universal Google requirement |
| Improve article structure | Reader can find the answer and its conditions | An ideal word or chunk length |
| Add supported structured data | Visible content matches valid markup | Accuracy of unverified claims |
Google’s browser-agent guidance discusses semantic interfaces and an evolving WebMCP proposal. Such integration work should be tested against the client that uses it. It is not evidence that every publisher needs a new protocol to appear in Google’s generative answers.
No new sitewide policy, ranking experiment or traffic result is announced here. Documentation can change; retain the review date and revisit the specific source before treating a claim as a continuing requirement.
For corrections, identify the claim, source and date when contacting Decentralised News. A new record-creation timestamp does not refresh its underlying evidence.
Change log: version 1.0.0, 2026-11-14. Initial documentation review, original instrument and labelled examples.
Frequently asked questions
Does this mean llms.txt is always useless?
No. A named service may use it. The point is that its usefulness to that service is a separate claim from being a Google Search requirement.
Should every long article be split into short pages?
No universal length follows from the reviewed guidance. Organize around the reader’s task and retain the conditions needed to interpret the answer.
Does correct schema guarantee an AI citation?
No. Syntax and eligibility do not guarantee indexing, display, ranking or citation. The visible content also needs to support the markup.
Is the inclusion setting the same as an AI-training control?
No. Google’s current documentation distinguishes its Search generative AI inclusion control from training controls. Check the intended system before changing either.
Does the checker inspect my website?
No. It evaluates entered claim scope against dated guidance. Any site condition is user-declared and needs separate verification.
Primary sources
- Google: optimizing for generative AI features. The five myth categories and the distinction between foundational SEO and unsupported AI-search prerequisites. Reviewed 2026-11-14.
- Google: AI features and your website. Indexing and snippet eligibility; no guaranteed inclusion. Reviewed 2026-11-14.
- Search Console: Search generative AI control. Current inclusion control, inheritance and separation from training controls. Reviewed 2026-11-14.
- Google: structured-data policies. Visible-content and structured-data consistency. Reviewed 2026-11-14.
- Google: spam policies. Boundaries for manipulative practices. Reviewed 2026-11-14.
- web.dev: Build agent-friendly websites. Optional agent-interface engineering and evolving WebMCP status. Reviewed 2026-11-14.
Turn the finding into a decision
A publisher evaluating a specific proposal can ask the DN Research Desk to scope a documented requirement and acceptance-test review. Provide the claim, target system and current evidence. Feasibility, access and a written quote come first. No ranking, traffic or citation outcome is guaranteed.
Related DN research
Disclosure: Analysis and instrument results are editorial. Where labelled affiliate links appear, DN may receive compensation when you sign up or buy through them. Commercial relationships do not determine findings. No tested integration or guaranteed outcome is implied.
