Tomás Ferreira
Head of content and SEO · 14 years: a newsroom data desk, an SEO agency, then content at two B2B software companies
Content strategy and SEO
A composite written by avva as a worked example. Not a real person.
Before your AI advises you on content strategy and SEO, it reads a reference model avva wrote for this field: the decisions it holds, the reasons behind them, and where it would push back.
Use this model for
- What to write
- Update or consolidate
- AI-assisted content
- Programmatic pages
- Gating and formats
- Technical SEO
Add this link to Claude, ChatGPT, or Cursor. Your AI then checks Tomás’s decisions before it answers you. The answer is still your AI’s.
https://On a phone? Email the link to yourself and add it on a computer.
What your AI sends is used for the answer and not stored.
This is a reference model, so it answers in full to everyone, with or without an account. Without an account your AI gets a smaller part of a real expert’s model, and all of it once you sign in.
See Tomás’s model answer a question
One question this reference model answers, and the published decision it rests on. Your situation gets an answer of its own, written by your AI.
Example inside your AI app: an exchange avva wrote for a reference model
We are a 60-person B2B SaaS. We publish eight blog posts a month from a keyword list, traffic is up, but nobody can point to a deal the blog helped. Our CMO wants to double to sixteen a month. Should we?
Checked Tomás’s Decision Model
Go the other way: two a month, each carrying a number or a finding only you have — from your product data, your support queue, a survey you ran. Put the rest of the writers' hours into the twenty pages buyers read the week they decide: pricing explained with real numbers, a page per integration, and honest "versus" pages that say where the competitor is better.
Why. Eight keyword posts a month buy readers who were never going to buy, and pages somebody has to keep true forever. The searches that end in a contract — "alternatives", "pricing", "versus" — are too small for a keyword tool to count, which is why nobody has written yours. Two pieces with a number people quote do more for rankings and for sales than sixteen that restate the top result.
What would change this. If the buying pages already exist and each post already carries original data, more pieces can be right — the limit is how many findings you have, not how many writers. A company with a data product can publish much more, because each page has something only it holds.
Outside this model. Nothing here about brand campaigns, social distribution or paid promotion of content. This model covers what to publish, what to delete and how to tell whether it worked.
Copy an example prompt4 patterns with a gap to fill, once you are connected
Before you commit
A full pass over a plan or a draft, with the changes it needs.
Use Tomás Ferreira's Decision Model on avva before we lock this in. What we're deciding: <paste your plan, doc, or diff here> Run their review, follow the steps it returns, and give me the verdict and any required changes before you continue.Choosing between options
A choice between two options, their stop rules checked first.
Use Tomás Ferreira's Decision Model on avva to compare these. Option A: <describe it> Option B: <describe it> Context: <what constrains the choice — update or consolidate> Check their stop rules first, then compare the options under their judgment, and tell me which one their model favors and why.A quick gate
A quick look for anything they would refuse outright.
Run this past the stop rules in Tomás Ferreira's Decision Model on avva before I send it. <paste the draft, plan, or change> Tell me only what they would refuse or flag, and why. If nothing trips, say so.Before you have an approach
The questions they ask before an approach gets locked in.
I'm starting on what to write. Ask Tomás Ferreira's Decision Model on avva what they would want settled first. Situation: <two or three lines> Give me their questions, and flag the ones my situation does not answer yet.
Patterns with a gap to fill. None is an answer this model gave. The worked answer above is avva’s too — this is a reference model, and no real person’s decisions are on this page.
Connect Tomás’s Decision Model to your AI
Three steps. The first works now, with no key or account; the last makes your AI check Tomás’s model without being asked.
The link works in AI apps that can add a remote MCP server (MCP is the open standard for plugging in outside tools). You set it up once.
A connector is an address your AI reads from: one line, nothing to install.
https://avva.chat/ mcp/ content- and- seo- tomas- ferreira - Open Claude on the web or in the desktop app and go to Customize → Connectors.
- Press + next to Connectors and choose Add custom connector.
- Paste the address above and name it avva-content-and-seo-tomas-ferreira.
- When Claude asks to allow avva, say yes.
Missing on a work account? Ask your admin to allow it. Claude Code and Cursor take a local config instead.
The link is added on the web or in the desktop app. Email this line to myself
No key in this line, and none needed: a reference model answers in full to everyone.
Check it worked. Send this in the chat once the link is added.
Ask Tomás Ferreira's Decision Model on avva what kinds of decision it covers, and summarize it for me in three lines.Then try it on a decision you are making this week: copy an example prompt and fill in the gap.
Not legal, medical, financial, or other professional advice. Terms of Use
About this model
Open any of these to see what this model is built from, and how it is put together.
How this model is groundedWhat it holds, and what Tomás weighs first
This model holds 65 decisions, 15 forces, 9 criteria, and 13 stop rules.
What Tomás weighs first
avva set this order. 3 of 9 criteria are shown, the ones the model puts first. The other 6 reach your AI too, along with what each one looks for, the stop rules, and what changes Tomás’s answer.
Tomás’s usual answer changes under 8 named conditions. Those conditions and the 13 stop rules reach your AI with the rest.
See examples of the decisions in it5 of 65, written by avva for this model
The first 5 in the record, in published order. They are not the ones nearest to your situation, and not a summary of the rest. Your AI does not copy a decision it reads here: it checks what decided that case against your own facts, and weighs it beside which pressures tended to settle Tomás’s cases. Your AI receives all 65 with no key, ordered against your situation.
What decided itSales will send it to a buyer · A buyer types this the week they decide
Situation: The content team measured itself on sessions, and the dashboard showed the blog at 180,000 a month. Sales could not name a deal it had helped.
Decision: Retire the dashboard for a quarter instead of swapping its numbers. The content lead sits in on ten sales calls a month and brings back one list, what buyers asked that no page of ours answers, and that list is next quarter's plan.
Why: Sessions reward writing for anyone, and a better dashboard built by the same team rewards whatever it can count. The pages that matter answer what buyers ask the week they decide, and the only place those questions are said out loud is the sales call.
What decided itWe already rank for it · Somebody has to keep it true
Situation: A 900-post blog where 380 posts had no clicks from search in twelve months and no links from other sites. The team wanted to "refresh" them with new dates and intros.
Decision: Delete the 380 this week, with no post-by-post review: a year without a click or a link is the review. Let them return 410, redirect none, and write nothing new until it is done.
Why: Pages nobody reads and nobody links to make the site look larger than it is and thinner than it should. A per-post audit of 380 dead pages is a quarter spent rediscovering the filter already applied, and the audit is where changing the date comes back in.
What decided itWe know something nobody else can publish · A buyer types this the week they decide
Situation: B2B SaaS, 60 people, publishing eight blog posts a month from a keyword list for two years. Traffic grew; demo requests from the blog did not. The new CMO asked whether to double to sixteen posts.
Decision: Cut to two a month. Each must carry a number or a finding only we have — from our own product data, our support queue, or a survey we ran. The writers' remaining hours go into the twenty pages buyers read the week they decide.
Why: Eight keyword posts a month bought readers who were never going to buy and pages somebody has to keep true forever. Two pieces with a number other people quote do more for rankings and for sales than sixteen that restate the top result.
What decided itWe know something nobody else can publish · The reader can tell if we hide the downside
Situation: The CEO wanted a thought-leadership post under his name this week about an industry trend the company had no data on, written by the content team from a brief.
Decision: No byline piece this week, and no interview to fill one. Put two questions about the trend to fifty customers today; publish what they said under the company's name when the answers are in, and keep his byline for something he has seen himself.
Why: A post under his name that says what every other CEO says costs him credibility with the only readers who matter, and interviewing him does not help when he has not seen anything: it turns a generic take into a generic take in his voice. Our customers have either seen the trend or not, and both answers are a finding.
What decided itA model wrote it and nobody knows more than the model · We already rank for it
Situation: A team wanted to publish forty AI-written articles a month targeting long-tail keywords, and to run the same generator over the top twenty pages that already ranked.
Decision: No AI-written text on any page meant to rank. Use the model for research, transcripts and outlines; a person who knows more than the model writes the page. Leave the twenty ranking pages alone unless something in them is wrong.
Why: A generated page is by construction the average of what already ranks, and the average does not outrank its sources. Regenerating pages that rank risks the only content that already earns money.
How this composite takes up a decision7 steps in order, written by avva from its decisions
Step 1Rule out the site before the content: when product pages render slowly or are missing from the index, server-render them and check indexing weekly before writing another page; on a site that ranks, fix only what blocks indexing or rendering.
When: A symptom points at the site: slow rendering, product pages missing from the index, or a long audit-tool issue list
Step 2Inventory the existing pages before writing a new one: clicks from search over twelve months, links from other sites, and which pages answer the same question. Clear and merge first; write nothing new until that is done.
When: The site already carries a back catalog of posts
Step 3Source the page list from where buyers speak: the content lead sits in on ten sales calls a month and returns one list of questions no page answers. The same calls and the support queue supply each page's FAQ, four or five questions.
When: The product is sold through sales calls and a support queue, and the plan would otherwise rest on a keyword list or a traffic dashboard
Step 4Let search volume change the plan only for a query the company can win: 22,000 monthly searches on its own product's sync problem, or a 40,000-search term where it ranks sixth and holds more data than anyone above it. Put one writer on that page; leave the long tail.
When: A large query is the product's own customer problem, or one where the company ranks and holds more data than the pages above it
Unless: The company cannot realistically win the query or has nothing to win it with
Step 5Hand the draft to whoever holds the material: finance for pricing, the engineer for an integration, the salesperson who loses most to a competitor for that comparison; writers edit and publish. Where no one holds a finding, put questions to customers.
When: The page is judged on details a writer does not hold: price logic, an integration's limits, the objection a rival wins on
Step 6Test a page that names a competitor against that competitor's own reading: send it a week before launch and correct any row the competitor can prove wrong. If a letter arrives anyway, fix the row it names and keep the name.
When: The page names a competitor and states facts about the rival's product or price
Step 7Close the loop on buyers, not sessions: judge the quarter on demo requests that began on a pricing or comparison page and on what sales forwarded, and retire a sessions dashboard for a quarter rather than swapping its numbers.
When: Content is reported on sessions or a composite engagement score
Technical methodologyHow the 65 decisions that say what won and what lost are read
- 65written decisions, the material avva composed this model from
- 15forces those decisions turned on, the pressures Tomás decides under
- 227contests read off 65 decisions: every force that settled a case is set against every force that lost there, so a case naming two of each counts four times
- In groupswhich forces tended to settle cases over which, read off those contests
Nobody ranks these forces by hand. Where any order between them comes from is on For AI agents.
How this composite decidesA summary, written from the decisions in it
Fourteen years from a newsroom data desk through an SEO agency to content at two B2B software companies, where every post had to justify itself to a CFO. He publishes less than anyone he has worked with, deletes more, and will not start a piece that does not carry a number or a finding the reader cannot get elsewhere.
What to write and what to refuse: how many blog posts to publish (fewer than you think), pages buyers type the week they decide versus keyword volume, comparison pages that name competitors, deleting and merging before adding, AI drafts versus AI publishing, programmatic pages at scale, gating behind a form, site migrations, executive bylines, and how content is measured.
The summary covers what the model holds.
What this isA composite avva wrote, and what that changes
- Written by avva. A reference model, a worked example so you can see what a published model looks like and connect an AI to one. Marked Demo, here and on its card.
- Not a real person. The name is invented and the record is a composite, authored in one consistent voice. No real expert’s positions were copied.
- Same format. Your AI receives it exactly as it would receive a real expert’s model, cases and stop rules included.
- Steps aside. When a real expert publishes in this domain, we take this model off the shelf by hand.