GEO Readiness Checklist
GEO Readiness: 28 Checks and a Prioritized Worklist
28 checks across crawl access, brand identity, applicable structured data, and useful content. Record evidence, distinguish optional settings, and prioritize corrections; this worklist does not predict AI citations.
v1.0 · June 2026 · Tenten GEO Research
CH.01
28 Checks to Organize the First Audit Week
Start with evidence about access, identity, markup, and content. These 28 checks support Tenten’s first week of a 30-day audit; they are a planning framework, not a sequence every AI engine follows.
A blocked search crawler can prevent it from fetching a page. That does not mean every AI system loses access or citations fall to zero. Training, search, and user-triggered requests have different purposes and controls. Identity information, schema, and clear writing can be checked alongside access; none guarantees citation.
Use a browser, curl, and relevant validation tools to document each check. Record pass, fail, unverified, or not applicable, with the URL, evidence, and date. A local response or a valid field does not by itself prove indexing, retrieval, or citation.
| Zone | What self-assessment can do | What deeper investigation can add |
|---|---|---|
| Crawl access | Inspect robots.txt, response status, and raw or rendered content; llms.txt is optional. | Check authenticated crawler requests and blocking in server/CDN logs. |
| Brand identity | Compare accurate names, descriptions, and existing relevant profiles. | Sample brand answers across six engines, recording model, date, locale, and sources. |
| Structured data | Check applicable types, syntax, and agreement with visible facts. | Investigate supported Search features and errors; do not infer schema’s citation effect. |
| Useful content | Check clear answers, sources, dates, and buyer questions. | For a six-engine baseline, define the 100–300 prompts, repeated observations, and denominators. |
Self-checks and deeper measurement answer different questions; neither reveals an engine’s hidden ranking weights.
Chapter takeaway
The output is an evidence-backed worklist. Passing all 28 checks is neither citation eligibility across engines nor a return-on-investment forecast. Optional or inapplicable items are not failures.
CH.02
Zone 1 · Crawlability: Can AI Get In? (7 items)
Separate the access policy you intend from accidental blocking. Test the relevant crawler and URL; one missing optional file does not make an entire site invisible.
| Check | How to test it | What the finding means |
|---|---|---|
| C1 · Crawler policy matches its purpose | Compare robots.txt with your intended training/search policy. GPTBot is not OAI-SearchBot; Google-Extended does not control Google Search. Check user-triggered exceptions in provider documentation. | An unintended block can limit the affected crawler’s access; it does not prove all citations disappear. |
| C2 · Check CDN/WAF responses and real requests | Use curl -A "GPTBot" as a diagnostic request, then inspect logs and verify crawler identity using provider IP/DNS guidance. A copied User-Agent is not authentication. | A 200/403 in a local test is evidence about that request, not proof of actual crawler visits or Search inclusion. |
| C3 · Important content is accessible | Compare raw HTML and the rendered page. Google renders JavaScript; not every bot does. Server-rendered text can broaden access without being a universal citation requirement. | Rendering or resource errors can hide content from a particular system; test it rather than declaring all client-rendered content nonexistent. |
| C4 · Measure response performance | Check TTFB with PageSpeed Insights or curl -w and investigate slow responses. Set a performance target for your site, not an AI citation threshold. | Slow responses can affect users and fetching; this checklist does not establish a fixed indexing delay. |
| C5 · Discovery URLs are maintained | Inspect sitemap.xml, recent URLs, HTTP status, and Search Console indexing evidence. A sitemap helps discovery but does not guarantee indexing. | Broken or omitted URLs warrant investigation; engines may discover pages through other links. |
| C6 · Indexing directives are intentional | Sample 10 important pages for noindex and canonical directives; compare them with the intended indexable URL and Search Console evidence. | Unintended directives can affect indexing or canonical selection; a canonical is not a promise of citation. |
| C7 · Review llms.txt only if you maintain it | If you use the optional proposal, check its links and intended consuming system. Mark absence not applicable. | Google Search ignores llms.txt: it neither helps nor harms visibility or rankings. Other systems’ use must be verified separately. |
Seven access checks, with optional files separated from actual fetching or indexing problems.
A WAF configuration and robots.txt can tell different stories. Inspect actual requests, status codes, and blocked resources before assigning a cause. Spoofing a crawler’s User-Agent can help diagnose a rule, but verified logs are needed to establish who visited; do not turn a hypothetical incident into a Tenten client result.
Reading this zone
Prioritize verified access problems according to the affected pages and systems. Optional llms.txt absence and an intentional training opt-out are not emergencies; a failed local test alone does not establish a citation loss.
CH.03
Zone 2 · Entity Building: Does AI Recognize Who You Are? (7 items)
Accurate brand identity helps readers check who is speaking. Existing profiles and independent sources can support that check; a QID, knowledge panel, or identical wording is not a universal AI citation prerequisite.
| Check | How to test it | What the finding means |
|---|---|---|
| E1 · Brand names describe the same organization | Compare official names, legal names, and legitimate localized variants across the site and relevant profiles; explain differences where needed. | Inconsistent or wrong names can confuse readers; spelling variants alone do not prove multiple engine entities. |
| E2 · Descriptions are accurate and consistent | Compare the About page and relevant profiles, including Google Business Profile only where applicable. Keep the meaning aligned, not necessarily identical sentences. | Outdated descriptions can misrepresent your business; record the actual inconsistency. |
| E3 · Existing knowledge-base records are eligible and sourced | Check existing Wikidata/Wikipedia records if relevant. Follow each community’s eligibility, reference, paid-editing, and conflict-of-interest rules; absence can be not applicable. | A QID is an identifier, not a verified badge or citation guarantee. Do not create unsupported records to pass this checklist. |
| E4 · Relevant Google business information is accurate | Where a Business Profile or knowledge panel exists and is relevant, check facts and official links. Do not require every business to have both. | Incorrect information needs correction; absence does not mean two engines are automatically lost. |
| E5 · Independent mentions are attributable | Inventory relevant third-party mentions, their sources, dates, and factual accuracy. Do not buy or fabricate corroboration. | Missing coverage is a sourcing decision, not proof that a brand can never be recommended. |
| E6 · Responsibility and expertise are visible | Check appropriate author/editor attribution, qualifications, and supporting profiles for core content; apply the context of the subject. | Missing attribution makes expertise harder to assess; this is not a universal engine penalty formula. |
| E7 · Branded results are understood | Sample the first 10 branded results and record wrong, stale, positive, or critical information. Seek corrections to factual errors without claiming control over independent reviews. | Use observed inaccuracies as follow-up work; critical coverage alone does not prove a lost buyer or citation. |
Seven identity checks: preserve truthful variants and mark irrelevant profile requirements not applicable.
Access fixes and identity corrections can proceed in parallel. Their publication, recrawling, and observed use are separate facts, with no fixed absorption period. In Tenten’s 90-day validation workflow, record identity work alongside content rewrites rather than promising an engine response date.
Reading this zone
Ask six engines about the brand only as a dated sample. Record sources, model, locale, and variability; an inaccurate answer identifies a problem to investigate, not a calculation that 3 of 7 identity checks failed.
CH.04
Zone 3 · Applicable Structured Data (7 checks)
Structured data can describe visible facts and support eligible Search features. Google does not require special schema for generative AI search. Valid markup is neither a citation prerequisite nor a ranking guarantee.
| Check | How to test it | What the finding means |
|---|---|---|
| S1 · Organization data reflects the real organization | Check applicable Organization properties and genuine identity links such as sameAs. Use schema.org validation and current Google guidance; do not invent missing facts. | Missing optional properties are not an automatic trust failure; incorrect identity data needs correction. |
| S2 · Article markup suits the page | For 10 article samples, check appropriate Article/BlogPosting data, real attribution, and truthful publication/update dates. Not every page is an Article. | A date field does not prove freshness or a higher rank; keep marked-up facts aligned with the article. |
| S3 · Visible Q&A and any markup agree | Check that useful questions and answers are visible and accurate. FAQPage is optional vocabulary, not an AI requirement; Google FAQ rich results ended May 7, 2026. | Missing FAQPage does not establish a citation loss. Do not mark up hidden or invented answers. |
| S4 · Navigation and breadcrumbs agree | Where BreadcrumbList is used, compare its URLs and hierarchy with the page’s real navigation. | A mismatch can misdescribe navigation; absence is not proof that topic authority cannot form. |
| S5 · Service/Product types and properties are appropriate | Use the applicable type and its documented properties. For example, provider belongs to Service; do not assume every Product uses the same fields. | Incorrect type or unsupported facts need correction; they do not prove competitors will replace your offering. |
| S6 · Markup matches visible facts | Compare schema with what readers can see, including prices, dates, attribution, and claims. Remove hidden, stale, or misleading statements. | Google structured-data violations can affect rich-result eligibility; they are not a demonstrated universal domain-wide citation penalty. |
| S7 · Validate the relevant format and feature | Check syntax with validator.schema.org and supported features with Rich Results Test/Search Console reports. Assess warnings in context, not as automatically harmless. | Successful validation does not guarantee a rich result or AI citation; unsupported features need separate treatment. |
Seven markup checks: applicability, accurate visible facts, and feature-specific validation.
Google’s guidance distinguishes structured-data eligibility from generative AI inclusion. Do not claim that all six engines consume every schema field or that markup universally improves retrieval. Use accurate markup for the supported purpose, and measure any observed changes separately.
- →Correct misleading facts and parse errors first; choose Organization, Article, or other types according to the page, not an invented citation-impact order.
- →validator.schema.org, Rich Results Test, and Search Console answer different questions. A missing unsupported-feature report is not a failed deployment.
- →After a major template change, check that applicable markup still matches the visible page.
CH.05
Zone 4 · Useful, Verifiable Content (7 checks)
These are editorial checks for people seeking an answer. They are not a proven highest-failure zone or an engine’s mandatory extraction format. Keep necessary context, evidence, and qualifications.
| Check | How to test it | What the finding means |
|---|---|---|
| A1 · Readers can find the answer | Review 10 core pieces and their first 100 words as a Tenten editing sample, not an engine threshold. Move the useful answer earlier when the topic permits. | A buried answer can slow readers down; there is no documented fourth-paragraph citation cutoff. |
| A2 · Headings help readers navigate | Compare headings with real buyer questions. Use question-form headings where helpful, without forcing an exact match to every prompt. | Generic headings may be less informative; Google can understand meanings and synonyms without exact prompt wording. |
| A3 · Evidence and original contribution are clear | Identify firsthand observations, original data, or useful analysis; name sources and methods. A careful synthesis can also serve a reader. | Unattributed or unsupported claims need work; lack of original data does not prove engines will never cite a page. |
| A4 · Definitions are understandable | Explain core terms clearly, with qualifications and references where needed. One paragraph is an editing option, not a citation rule. | An unclear definition can confuse a reader; this worklist does not establish definition-prompt volume or traffic loss. |
| A5 · Commercial facts are clear where appropriate | Check readable specifications, pricing, and conditions that the business can disclose. Explain custom-quote terms instead of inventing a public price. | Missing or ambiguous facts can hinder comparison; disclosure needs depend on the offering and obligations. |
| A6 · Dates reflect substantive updates | Review whether claims and sources remain accurate. Tenten may use 12 months as a review reminder; change dateModified only for a substantive update. | Age alone does not prove citation decay. Correct obsolete facts without refreshing dates cosmetically. |
| A7 · Buyer questions receive useful answers | Check relevant high-intent questions on service pages; use an FAQ only when it helps the reader. | Unanswered questions are editorial gaps, not a measured universal purchase-intent or citation-rate penalty. |
Seven editorial checks, with recommendations separated from observed engine outcomes.
Reading this zone
Use the gaps you actually found to order edits. Access does not guarantee citation, and this checklist does not rank the four zones by an undocumented internal failure rate. Measure changes instead of attributing every missing citation to content format.
CH.06
From Evidence to a Prioritized Worklist
Keep the 28 checks as records, not a 28-point grade. Mark pass, fail, unverified, or not applicable, with evidence and a responsible next action. There is no validated score here that predicts citations or commercial return.
STEP 1
Record the result and evidence
Work through CH.02–CH.05. Add URL, test date, method, and what was observed. A validator pass proves its checked condition, not engine inclusion; leave missing evidence unverified.
STEP 2
Distinguish applicability and severity
Do not count an optional llms.txt file, absent eligible knowledge record, or irrelevant schema type as a failure. Prioritize the affected pages and real consequences, not equal-weight subtotals.
STEP 3
Assign the next action
Use the three priorities below as Tenten’s planning guidance. Give each action an owner and a review point; this is not an industry certification or an engine’s grade.
| Observed issue | Priority | Next action |
|---|---|---|
| Verified blocker | Restore intended access or indexing | Fix the affected crawl, response, rendering, or indexing problem and verify its resolution. Pause only work that depends on that specific blocker. |
| Identity or content gap | Correct facts and clarify the answer | Advance applicable identity, schema, and editorial corrections in parallel. A 3-week diagnose → rewrite → verify loop is Tenten’s working cadence, not an engine deadline. |
| Measurement follow-up | Establish an observed baseline | With a defined prompt library, record repeated citation/mention observations, dates, locales, models, and denominators. Keep pipeline attribution separate from engine presence. |
Three action priorities, without numerical readiness bands or predictive grades.
In Tenten’s planning workflow, the first week of a 30-day audit documents these checks; Weeks 2–4 establish a prompt baseline and investigate gaps. A 90-day validation period can use 3-week rewrite loops and monthly pipeline reviews. A self-check is useful preparation, not completion of a professional audit or confirmation of a revenue forecast.
One last line
Keep a short list of verified problems, unanswered questions, and the next responsible action. That is enough to start useful work without inventing a score or an expiring opportunity.
Sources and methodology
Tenten authored this 28-item, four-zone worklist as planning guidance, not a validated citation score or a published client study. The former per-zone failure chart is retired because its sample, period, coding, and underlying records have not been established here; its figures are not declared false. Sources below support specific access, optional-file, schema, identity, and substantive-date distinctions, not guaranteed citations. Google’s May 8, 2026 update states FAQ rich results ended May 7; its June 15 update says llms.txt has no positive or negative Google Search visibility/ranking effect. Provider rules differ; verify the relevant system. Reviewed with substantive corrections on 2026-10-05. https://platform.openai.com/docs/bots · https://developers.google.com/search/docs/fundamentals/ai-optimization-guide · https://developers.google.com/search/updates · https://developers.google.com/search/docs/appearance/structured-data/sd-policies · https://developers.google.com/search/docs/appearance/publication-dates · https://developers.google.com/crawling/docs/crawlers-fetchers/verify-google-requests · https://www.wikidata.org/wiki/Wikidata:Notability
© 2026 Tenten GEO · You're welcome to cite this checklist with attribution.
Discuss your AI search priorities
Request a 30-minute benchmark interpretationSend a booking request. We’ll email you to confirm a time within 24 hours.