inemaSEO 2 — Content: pages only you can write|INEMA.CLUB-PRO

Lesson 5 · 18 min

Listing pages with real data and the anti-fabrication check

A catalog of products, courses, or services becomes one page per item: the listing page. Built in bulk, with template text, it is the thin page that doesn't get indexed. Built from each item's real page, with an AI that only extracts and a check that discards whatever it made up, it becomes content only you have.

Start the lesson
An archivist checks a catalog listing page against the original document, in front of a wooden filing cabinet.

In this lesson you will

  1. 1Why 283 template-based listing pages didn't get indexed
  2. 2The extractive prompt: copy from the source, don't create
  3. 3The check: title, words, and numbers verified against the source
  4. 4Publish a little at a time, and only what passed

Goal: generate catalog listing pages extracted from each item's real page and discard any item that fails the anti-fabrication check.

01From the real page to the published listing page

Tap each block.

Click each block in the diagram to see what it does.

The AI rewrites and organizes; the check makes sure it did not make up facts.

02The case and the fix

Tap each step.

1. The batch

INEMA published 283 course listing pages in a single commit. Of the 6 listing pages in the sample, 5 were "discovered, not indexed" and had never been crawled.

03The extractive prompt and the check

Replace "course" with your own type of item. Run the check on every sentence and title.

prompt-ficha-extrativa.md (inemaSEO templates)
You extract a course's syllabus from the TEXT OF ITS PAGE. Rules:
- EXTRACTIVE: copy module/track titles exactly as they appear (you may remove emoji). "Audience", "learns" and "builds" phrases
  must reuse the words on the page; do not invent anything.
- Do not include price, free access, promotions, testimonials or student counts.
- If a field is not on the page, use [] or null.
Answer ONLY JSON: {"estrutura":[{"titulo":"","detalhe":""}],"aprende":[""],"publico":[""],
"constroi":[""],"pre_requisitos":[""],"carga_horaria":null,"nivel":null}
Limits: estrutura up to 12 items — only tracks/modules/lessons of the content, never sections of the page
("Who it's for", "FAQ", "Start now").
Run the check (validar_extracao.mjs)
git clone https://github.com/inematds/inemaseo-metodo.git
cd inemaseo-metodo
# fonte.txt = text of the item's actual page
node tools/validar_extracao.mjs fonte.txt "Hydration and folds"
node tools/validar_extracao.mjs fonte.txt "Natural-fermentation bread course with 12 lessons"
fonte.txt used in the example
Natural-fermentation bread course
Module 1: The levain from scratch
Module 2: Hydration and folds
Module 3: Home oven
Workload: 6 lessons of 20 minutes each.
For people who already bake bread with regular yeast and want to switch to levain.
Example from the repository
$ node tools/validar_extracao.mjs fonte.txt "Hydration and folds"
{ apoiada: true, numerosNaFonte: true, tituloNaFonte: true }

$ node tools/validar_extracao.mjs fonte.txt "Natural-fermentation bread course with 12 lessons"
{ apoiada: true, numerosNaFonte: false, tituloNaFonte: false }

$ node tools/validar_extracao.mjs fonte.txt "Learn how to sell your breads on Instagram"
{ apoiada: false, numerosNaFonte: true, tituloNaFonte: false }

04Practice: three real listing pages

Pick 3 items from your catalog that have their own page with text.

E1 · Extract three listing pages

40 min

Save each item's page text to a file and run the extractive prompt with the agent from your subscription.

Done when: you have 3 JSON files in the prompt's format, with no price, testimonials or customer counts.

E2 · Run them through the check

30 min

Run validar_extracao.mjs on each title and sentence from the 3 listing pages. Discard whatever fails.

Done when: a note lists each item with its check result, the rejected ones are out, and only listing pages with 3 or more valid modules were marked to publish.

05Test what you learned

The explanation shows up after your answer.

Question 1 of 4

The check rejected a sentence because the number is not in the source. What should you do?skill

06Takeaways

Real sourceEach listing page comes from the page of the item itself.
CheckTitle, words, and numbers are checked; whatever fails is dropped.
In batchesPublishing hundreds at once is what stalled the crawl.

Going further

validar_extracao.mjs (inemaSEO, MIT)prompt-ficha-extrativa.md (inemaSEO, MIT)Google: spam policies, scaled content (English)
Previous lessonNext lesson