Lesson 2 · 20 min
The diagnosis is the first deliverable and the entry point of the sale. Three parts, all read-only: what the server delivers to the search engine, how ready the site is for AI and agents, and whether the client shows up when someone asks the AIs about their niche. It all fits on one page.
Start the lesson
Goal: put together a site's diagnosis in three parts (delivered HTML, AI readiness, neutral questions in the AIs) and present it on a call.
Tap each block.
Everything is read-only. Changing the site or opening the client's dashboard requires written authorization (lesson 7).
Tap each step.
Run the method's auditoria.sh or the inemaseo-sistemas auditor on the public site, with a pause between pages.
Separate what is important (a page with no text in the HTML, wrong canonical) from what is an improvement (long title).
On 09/30 the course index became "search only" and 310 pages lost the only internal link the server delivered. The HTML X-ray is what catches this kind of error.
The score only counts if the weights are public: the client sees why they lost points.
It measures readiness, not a guarantee of being cited. Say so in the report.
Build 20 questions with the client that their audience would ask, without mentioning the brand. Freeze them as v1: changing them later breaks the comparison.
Run them by hand and note the results in a spreadsheet. Answers vary: it's a sample from that day, repeated every month.
On 07/10/2026 INEMA showed up in 0 of 48 neutral answers on Claude and 0 of 4 on Gemini. And "What is INEMA?" brought back 4 of 4 answers about another organization with the same name: a name collision.
Three findings with the why, the question set summary, and one next step.
The detailed teaching of the tools is in course 7 (Systems).
Open tools (MIT). Replace what is between < > with your own. Read-only, on a public site.
git clone https://github.com/inematds/inemaseo-metodo && cd inemaseo-metodo tools/auditoria.sh https://www.<client-site.com> /<page-1>/ /<page-2>/
git clone https://github.com/inematds/inemaseo-sistemas && cd inemaseo-sistemas python3 ferramentas/auditor/auditor.py https://<client-site.com> --max 50 python3 ferramentas/agente-ready/agente-ready.py https://<client-site.com>
date,assistant,question,run,client_cited(yes/no/partial),competitors_cited,sources_cited,factual_errors <YYYY-MM-DD>,<assistant>,<neutral question 01>,1,,,,
DIAGNOSIS — <site> — <date> 1. What the search engine sees (X-ray) - Finding 1 (important): <what> → <why it matters> → <what to do> - Finding 2: ... - Finding 3: ... 2. Readiness for AI and agents: score <N>/100 (public weights). Lost points on: <...> The score measures readiness; it does not guarantee being cited. 3. Questions in the AIs (question set v1, <date>, by hand): showed up in <x> of <y>. Who shows up instead: <...>. Sources cited: <...> 4. Non-branded number (if there is access): <clicks>/<impressions> in 28 days 5. Next step: 30/90/180 plan (lesson 3)
| priority | finding | pages |
|---|---|---|
| important | Little text in the delivered HTML (fewer than 80 words) | 2 |
| improvement | Title outside 15–65 characters | 28 |
| improvement | Organization/Person/WebSite without @id | 1 |
Crawl of 50 pages, GET only, 1 s pause. The "info" rows (for example, FAQ markup, whose rich result Google stopped showing on 05/07/2026) are left out of the client report.
| criterion | weight | result |
|---|---|---|
| search bots allowed | 12 | ok |
| CDN doesn't block bots | 12 | ok |
| text without JavaScript | 10 | ok |
| authorship | 5 | partial |
| dates | 5 | partial |
| about/contact page | 2 | fail |
Score 93 out of 100 (pesos.json v1.0.0). Real examples in inemaseo-sistemas/exemplos.
Simulated conversation for practice. Choose the answer you would give; the comment appears right after.
Choose your answer. The feedback appears right after, and the conversation continues based on your choice.
Use your own site or one belonging to someone who authorized it in writing.
Run the auditor (or auditoria.sh) and agente-ready on a site of yours. Pick the 3 findings that weigh most and write, for each one, why it matters and what to do.
Done when: you have the output files from both tools and a list of 3 findings with "why it matters" and "what to do", each with the affected URL.
Write 20 neutral questions the site's audience would ask an AI, without the brand name. Run 5 of them by hand in 2 assistants and log them in the spreadsheet.
Done when: the spreadsheet has the 20 questions marked as v1 and at least 10 rows filled in (5 questions × 2 assistants) with the date and "cited yes/no".
The explanation appears after your answer.
Question 1 of 4
Why can't the question set's questions mention the client's name?skill
Question 2 of 4
The site's readiness score came out at 93 out of 100. What do you tell the client?
Question 3 of 4
For the diagnosis, what access do you need?
Question 4 of 4
Why freeze the question set as v1?
Choose an option. The explanation appears after your attempt.