E1 · The score
15 minRun agente-ready on your site and write down the score and the 3 failing criteria with the highest weight.
Done when: you have the score, the 3 criteria with the weight of each, and the evidence the report shows.
Lesson 3 · 16 min
agente-ready: a 0–100 score with weights you can change
"AI readiness" scanners give you a score, but almost none publish how it is calculated. agente-ready keeps each weight, the reason, and the source in a versioned file, pesos.json. You get your site's score, see what weighs most, and if you disagree with a weight, you change it and run again.
Start the lesson
In this lesson you will
- 1The 6 categories and the weight of each criterion
- 2How the score is calculated (and what leaves the denominator)
- 3Why llms.txt weighs 2 and WebMCP is preparation
- 4Run with your own weights
Goal: get your site's readiness score and change one weight in pesos.json to see the score change.
01How the score comes out
Weights add up to 100. Tap each step.
Click each block in the diagram to see what it does.
A score with the math in the open: you can disagree with a number and show why.
02The criteria that weigh the most
Tap each group.
1. Access
robots.txt lets search bots (12) and user-requested agents (6) in; the CDN does not block the User-Agent of OAI-SearchBot, PerplexityBot, and Perplexity-User (12).
If a bot can't read the page, nothing else matters to it.
2. Content
150+ words in the delivered HTML (10), a visible answer with a main heading and paragraph (6), authorship (5), and date (5).
Authorship and date are what event sites and landing pages lack most.
3. Entity
Organization or Person with @id in the home page JSON-LD (8), sameAs for the official profiles (4), and a link to About or Contact (2).
This is the "who publishes" that course 2 teaches you to build.
4. llms.txt and WebMCP
llms.txt weighs 2: in the Ahrefs study (137 thousand domains, 06/15/2026), 97% of the files received no requests at all.
WebMCP weighs 8 in total: a tool registered with document.modelContext.registerTool() is worth 5. The specification is a W3C community draft (10/16/2026); the old API name earns partial credit, with a warning.
2026 yardstick
WebMCP is preparation for agents, not a ranking factor. llms.txt is cheap and optional, not a citation lever.
03Get the score and change a weight
Run it from the ferramentas/agente-ready folder.
Terminal · your site's score
cd ferramentas/agente-ready python3 agente-ready.py https://yoursite.com --saida nota/yoursite
Terminal · your weights
cp pesos.json my-weights.json # open my-weights.json and change the "peso" of one criterion (e.g.: webmcp_api from 5 to 0) python3 agente-ready.py https://yoursite.com --pesos my-weights.json --saida nota/yoursite-my-weights
Why I changed the weight (template)
# Weights for <yoursite.com> — <dd/mm/yyyy> criterion: <criterion id> original weight: <n> → my weight: <n> reason: <e.g.: my audience doesn't use browser agents; WebMCP is a draft with no stable browser> score with pesos.json: <n> · score with my weights: <n>
Example from the repository
www.inema.club: nota 93/100 (5 páginas, 22 requisições). [falha, peso 2] Link para Sobre ou Contato na home: sem link /sobre, /about ou /contato eventos.inema.pro: nota 69/100 (5 páginas, 15 requisições). [falha, peso 8] Organization ou Person com @id no JSON-LD da home: 0 entidade(s); com @id: 0 [falha, peso 5] Autoria identificada: 0 de 5 com autor (meta, rel=author ou JSON-LD) [falha, peso 5] Data de publicação ou atualização: 0 de 5 com data (JSON-LD, meta article ou <time>) [falha, peso 5] Ferramentas WebMCP registradas com document.modelContext.registerTool(): nenhum registerTool() em 2 script(s) de mesma origem + inline
The same eventos.inema.pro with webmcp_api from 5 to 0 in my-weights.json:
eventos.inema.pro: nota 73/100 (5 páginas, 15 requisições). [falha, peso 8] Organization ou Person com @id no JSON-LD da home: 0 entidade(s); com @id: 0 [falha, peso 5] Autoria identificada: 0 de 5 com autor (meta, rel=author ou JSON-LD) [falha, peso 5] Data de publicação ou atualização: 0 de 5 com data (JSON-LD, meta article ou <time>) [falha, peso 4] sameAs ligando a entidade aos perfis oficiais: sem sameAs na entidade
The math: 66 points earned out of 95 applicable gives 69; out of 90, it gives 73. The list of what to fix first did not change order at the top.
- Treating the score as a ranking: the weights are an editorial decision by INEMA, with a source for the why, not for the number.
- Changing a weight to make the score go up: change it when you have a reason and write it down; the criterion stays in the report.
- Expecting to see a WebMCP tool from another origin: detection is static; a script from another domain, loaded later or in an iframe does not show up.
04Practice: the score and your weight
In lesson 7 the score joins the monthly routine.
E2 · Disagree with a reason
15 minCopy pesos.json, change one weight with a written reason, and run again with --pesos.
Done when: the "why I changed the weight" template is filled in with both scores, and they are different.
05Test what you learned
The explanation shows up after your answer.
Question 1 of 4
Your site has no form. What happens to the 4 points for forms?skill
Correct
A criterion that does not apply is left out of the math. The score is points earned divided by the applicable weights.Not that one
A criterion that does not apply is left out of the math. The score is points earned divided by the applicable weights.Review: How the score comes outQuestion 2 of 4
Why does llms.txt weigh only 2?
Correct
It is cheap and useful as an index for coding agents, but there is no evidence it leads to citation. Google's AI guide says it may ignore it.Not that one
It is cheap and useful as an index for coding agents, but there is no evidence it leads to citation. Google's AI guide says it may ignore it.Review: The criteria that weigh the mostQuestion 3 of 4
In the eventos.inema.pro output, what matters most for raising the score?
Correct
The output lists the failures by weight. llms.txt was already fine. Blocking training is informational and does not change the score.Not that one
The output lists the failures by weight. llms.txt was already fine. Blocking training is informational and does not change the score.Review: Get the score and change a weightQuestion 4 of 4
Does WebMCP registered on the site guarantee a citation in ChatGPT?
Correct
The specification itself says it is not a W3C standard. It prepares the site for agents in the browser; it is not a ranking factor.Not that one
The specification itself says it is not a W3C standard. It prepares the site for agents in the browser; it is not a ranking factor.Review: The criteria that weigh the mostQuiz complete
You got {a} of {n} right. Get them all right and the lesson is marked as complete.Pick an option. The explanation appears after you try.
06Takeaways
Open mathWeights, reason and source in pesos.json; whatever doesn't apply is left out of the denominator.
What weighs mostBot access (30) and extractable content (26) come before everything else.
2026 yardstickllms.txt is optional (weight 2); WebMCP is preparation, not ranking.