← Back

FRAMEWORK · v1.0

GEO Maturity Model v1.0.
Every criterion, in the open.

GEO maturity is a five-level scale (Read, Read correctly, Recommended, Cited, Driving action) that shows, on one line from externally visible security to people’s actions, how far a website has grown inside generative AI answers. KnownFor defines it and judges it with real numbers.

  • Version: 1.0
  • Published:
  • Publisher: KnownFor (株式会社アーグマックス)

GEO maturity is a yardstick defined by KnownFor, with every criterion published. It is not an official standard. As far as we know, there is no official standard for GEO maturity, such as an ISO or JIS standard (as of October 2026). So this page publishes exactly what we measure and where each pass line sits. Anyone can check the same criteria, even before signing up.

GEO (Generative Engine Optimization) is the work of getting named and cited inside generative AI answers. It is also called LLMO.

The five levels

Security and being chosen by AI
sit on one line.

The foundation is the security visible from outside, and AI being able to read you. On top of that come being read correctly, being recommended, being cited and driving action. Levels go up in order from the bottom. Passing a higher level early does not let you skip the ones below.

  1. Level1ReadSecurityAINo serious gaps in the security visible from outside. AI search crawlers can get in, read your text, and your pages are in the index.
  2. Level2Read correctlyAIIn answers that name you, prices, contact details and other facts match the facts you registered (90% or more).
  3. Level3RecommendedAIIn questions that don’t name you, the share of answers in which AI recommends you rises clearly from where you started, or stays at half or more.
  4. Level4CitedAIThe share of answers that use your pages as a source rises clearly from where you started.
  5. Level5Driving actionAIActionPeople come to your site from AI answers, and it leads to actions such as inquiries and sign-ups (when GA4 is connected).

Pass criteria

What it takes to pass each level.

Each level is decided by measuring fixed items. If every item passes, the level passes; if any item fails, the level is not yet passed. If an item has no source connected or no record, the level shows “no record” in gray and is called neither passed nor failed. We never fill in numbers by guesswork.

LEVEL 01

Level 1: Read

No critical gaps in externally visible security; AI search crawlers read the site without being refused; robots.txt does not block answer-use AI crawlers; page text does not depend on JavaScript; and the top page is in Google’s index.

ItemCondition and source
SecurityZero critical issues in the basic security items (https, where forms send, certificate expiry) in the most recent scan that includes security. Critical issues in email settings or component versions are shown in the alert level and not counted here. Source: the KnownFor scan (a passive check of what is visible from outside). No scan with security means “no record”.
AI search crawlersIn the last 7 days, AI search crawlers (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User, Amazonbot) read the site at least once without being refused. It does not pass if, among crawlers that came 3 or more times, 50% or more were refused on 50% or more of their visits (responses 401, 402, 403, 406, 429, 451, 503). Training crawlers (GPTBot, ClaudeBot and others) are not judged, because blocking them is a legitimate site policy. Source: Cloudflare (records of AI crawler visits). Not connected means “no record”.
robots.txtOn the pages checked, answer-use AI crawlers (Googlebot, Bingbot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User) are not blocked by robots.txt or Content-Signal (search=no, ai-input=no). Source: the KnownFor page check. No page check means “no record”.
Page textNone of the pages checked shows its text only after JavaScript runs (fewer than 200 characters of text in the first HTML, with scripts). Source: the KnownFor page check.
Google’s indexThe top page is in Google’s index (if the top page has not been inspected, at least one inspected page is). Source: URL Inspection in Search Console. Not connected means “no record”.
Bing’s indexNot part of pass or fail for now. KnownFor cannot check it yet, so it is listed in gray as “no record”.
The level passesAll five items above (excluding Bing’s index) pass.

LEVEL 02

Level 2: Read correctly

Among AI answers that named you in the last 28 days, at least 10 could be checked against your registered facts, and in at least 90% of them every checked fact was correct.

ItemCondition and source
What is comparedFacts stated in AI answers that named you in the last 28 days (Price, Audience, What you offer, What sets you apart, Area, Contact, Provider), against the facts registered in your fact ledger.
How answers are countedAn answer counts as correct only if every checked fact matches. One outdated or wrong fact makes the answer incorrect. Answers containing only facts you have not registered are not counted.
How facts are judgedAn AI (temperature 0) extracts claims from the answer with quotes; claims whose quote is not in the answer are discarded. Amounts, phone numbers, email, URLs and postal codes are compared by machine. Claims about a different organization with the same name are excluded.
The level passesAt least 10 answers could be checked, and 90% or more of them are correct. Fewer than 10 means “no record” (too few to judge).
SourcesAnswers from KnownFor AI checks, and your fact ledger (facts you registered).

LEVEL 03

Level 3: Recommended

For questions that do not name you, the range (Wilson 95%) of how often AI names you moves clearly above the baseline, or its lower end is 50% or more, with no group clearly falling below its baseline.

ItemCondition and source
What is comparedAmong answers to questions that do not name you, the share that names you. Source: KnownFor AI checks (Gemini, ChatGPT).
GroupsEach combination of product or service (unit) and AI (Gemini, ChatGPT) is compared separately.
Baseline and nowBaseline = the oldest AI check in the current run under the same conditions (same question version and same model). Now = the newest. If the question version or model changes, the baseline starts over. With fewer than two AI checks under the same conditions, the result is “no record”.
A group passesThe lower end of the current range (Wilson 95%) is above the upper end of the baseline range (moved clearly up), or the lower end of the current range is 50% or more (already strongly recommended).
A group falls clearlyThe upper end of the current range is below the lower end of the baseline range.
The level passesAt least one group passes and no group falls clearly. While the ranges overlap, it is “within normal variation” and does not pass.

LEVEL 04

Level 4: Cited

For questions that do not name you, the range (Wilson 95%) of how often AI cites your pages as a source moves clearly above the baseline, with no group clearly falling below its baseline.

ItemCondition and source
What is comparedAmong answers to questions that do not name you, the share that cites your pages as a source. Source: KnownFor AI checks (Gemini, ChatGPT).
GroupsEach combination of product or service (unit) and AI (Gemini, ChatGPT) is compared separately.
Baseline and nowSame as Level 3.
A group passesThe lower end of the current range (Wilson 95%) is above the upper end of the baseline range (moved clearly up). There is no 50% line as in Level 3.
A group falls clearlyThe upper end of the current range is below the lower end of the baseline range.
The level passesAt least one group passes and no group falls clearly. While the ranges overlap, it is “within normal variation” and does not pass.

LEVEL 05

Level 5: Driving action

In GA4, at least one visit from AI in the last 28 days, and at least one key event (such as an inquiry or sign-up) from AI visits.

ItemCondition and source
Visits from AIAt least one visit in the last 28 days whose GA4 source is an AI (ChatGPT, Perplexity, Gemini, Copilot, Claude, You.com, Phind, Poe).
Key eventsAt least one key event (such as an inquiry or sign-up) on landing pages of visits from AI. If key events are not set up in GA4, the result is “no record” (not zero).
The level passesBoth items above pass.
SourcesGA4. Not connected means “no record”.

Moving up and down

You move up only by holding a pass.

Results are calculated from real numbers every time, never set by hand. They are recorded per period of one week (7 days starting on Monday).

  1. PassWhen every item of the level passes, a passing run begins.
  2. Verification window (28 days)For 28 days from the start of the run, nothing is counted. This keeps a lucky result from moving you up.
  3. Hold 2 passing periodsWhen 2 passing periods come after the window, you move up a level. Periods that still rest on the same AI check count as one, so the same numbers are never counted twice.
  4. A fail resets the runIf a period fails along the way, the run breaks and starts over. Periods with no record are not counted, but they do not break the run.
  5. WatchingAfter you move up, we keep checking every week that you still hold the levels below. When AI answers change, a level you passed can slip. A slipped level is shown as “Slipped” and comes back as the first task to fix (the date you moved up stays on record).
  6. Maturity = the highest level reached in a rowYour GEO maturity is the highest level reached continuously from Level 1.

Data sources

What we measure with, and what we cannot measure.

SourceWhat it showsLevel
KnownFor AI checksAnswers from Gemini and ChatGPT to questions that do not name you and questions that do (model name and question version are recorded)2, 3, 4
Fact ledgerThe facts you registered2
KnownFor scanSecurity visible from outside (a passive check, with no intrusion or attack)1
KnownFor page checkrobots.txt, Content-Signal, text in the first HTML1
CloudflareVisits from self-identified AI crawlers and how many were refused (last 7 days)1
Google Search ConsoleGoogle’s index, via URL Inspection1
GA4Visits from AI (last 28 days) and key events on their landing pages5

What we cannot measure

  • We do not currently observe answers in Google’s AI Overviews or AI Mode. In GA4, visits from them cannot be told apart from ordinary Google Search either.
  • The AIs whose answers we observe are Gemini and ChatGPT. Answers from Claude, Perplexity, Copilot and others are not part of the level assessment.
  • We cannot check Bing’s index yet (it is not part of Level 1’s pass or fail).
  • AI crawlers are counted only when they identify themselves in the User-Agent and pass through Cloudflare.
  • AI answers change every time you ask. Shares are compared as ranges (Wilson 95%), and while the ranges overlap we call it normal variation, not a rise or a fall.
  • Fact matching includes judgment by AI (numeric items are compared by machine).
  • GA4 counts only people who clicked a link in an AI answer. People who saw your name in an answer and searched for you later are not counted.
  • A level describes where you stand now. It does not promise sales, search rankings or a place in AI answers.

For businesses and public bodies

One scale, two readings.

GEO maturity applies with the same five levels and the same conditions both to businesses, whose goal is to be chosen, and to organizations whose goal is to guide people correctly (local governments, medical institutions, police and so on). What changes is how you read Level 3 and Level 5.

Businesses

Level 3: RecommendedWhen asked “What is the best ___?”, AI names your company among the options.

Level 5: Driving actionPeople who come from AI answers go on to inquire, sign up or buy.

Local governments, medical institutions, police and others

Level 3: RecommendedFor residents’ questions like “Where is the nearest shelter?” or “My child has a fever at night. What should I do?”, AI points to the right page or contact of that organization.

Level 5: Driving actionPeople who come from AI answers go on to complete a procedure, check information or get in touch on the right page. Each organization decides in GA4 what counts as a key event.

Here, Level 2 “Read correctly” (are phone numbers, addresses and opening hours right?) matters most.

Versions

Versions and revision history.

Whenever we change the criteria (pass conditions, numbers or sources), we raise the version and record the date and what changed in this table.

VersionDateChanges
v1.0Published (first version)

Citing

Cite the source, and quote freely.

You are free to quote and introduce this page as long as you credit the source (KnownFor, “GEO Maturity Model v1.0”, and the URL).

ExampleKnownFor, “GEO Maturity Model v1.0” (October 27, 2026), https://www.knownfor.jp/en/maturity/

See your level, in real numbers.

A KnownFor plan (KnownFor Weekly, KnownFor 3x a week or KnownFor Daily) keeps track of your GEO maturity. Every plan measures all Levels 1–5 and shows you one step at a time what to do to reach the next level. The plans differ in how much you can use; with weekly observation, KnownFor 3x a week and KnownFor Daily reach a verdict sooner.

Sign up About KnownFor

The GEO Maturity Model is a yardstick set by KnownFor (株式会社アーグマックス). It is not an official standard. Levels are assessed in the KnownFor member screens.