38 SEO pages in a day, and a validator that rejects fake numbers

Today BuzzyX gained three new sections: reply examples, growth guides and post ideas. Thirty-eight pages, of which 33 are published and 5 are drafts. This is a note on how they were built and what we are testing, written before any of them has been indexed, so nothing here is a result.

Why these pages

The product is for people who want to grow on X without gaming it. The things those people search for are not "X growth tool". They search for what to reply when someone compliments them, how to answer criticism without getting defensive, what to post when they have nothing to say, how often to post. If the honest answer to those questions is useful on its own, it is worth publishing whether or not the reader ever signs up, and some of them will.

That last condition is the rule for every page: it has to be useful to someone who never signs up. Otherwise it is a doorway page, and those do not deserve to rank.

One template, many content files

Each page is a content file, not a hand-built page. The file holds a title, a description, an opening, sections made of typed blocks, example cards, a FAQ, topic tags and a few hand-picked related links. One template renders all of them: the centred title and byline, a sticky table of contents with reading progress, the example cards with copy buttons, a product card after the first section, a FAQ with structured data, and related pages chosen by shared tags.

Adding a page means adding a file. The sitemap, the hubs, the footer and the related links update themselves. A new category needs one small route; a new page needs nothing but content.

The validator

This is the part I wanted to write about. Every page passes through a validator before it can be published, and the same validator runs in the test suite, so a bad page fails the build. It checks:

• No em dashes or en dashes anywhere. A house rule, and the thing AI-written text leaks most.
• Unique titles, descriptions, headings and target keywords, so two pages cannot compete with each other.
• Length floors: at least 600 words for a published page, and most run between 1,500 and 2,000.
• A minimum number of example cards for the categories that promise examples, and a hard 280-character limit on reply examples.
• No unsourced statistics. Phrases like "studies show" or "40% more engagement" fail the page. If a number is not from our own data or a named source, it does not go in.
• No promised outcomes. "Guaranteed" fails the page.
• Every internal link points at a page that exists and is published. You cannot link to a draft.
• No repeated example text across pages, and a check for near-duplicate pages: two pages that share more than a fifth of their six-word phrases fail, and more than about a twelfth gets flagged for a human to read.

That last check is the one that stops the usual programmatic SEO failure: fifty pages that are the same page with the profession swapped. The audience variants on this site, for founders, developers, designers, marketers and creators, had to be organised around the situations each of those people actually meets, with examples only that person could write, or the validator would not have let them through.

What is marked as unverified

The pages were chosen from a keyword list with a column for evidence of demand. Right now every row in that column says "unverified". We have no Search Console data yet, and no keyword tool. Each query is there because it is something a real person plausibly types, with a note on why we think so. When the pages have been indexed for a few weeks, the real impressions replace the guesses and the priorities get re-sorted.

The five drafts are pages whose intent overlaps a published page: reply examples for indie hackers next to the founders page, a marketers guide next to the pillar, a templates page that risks being a template dump. They stay drafts, returning a 404 in production, until data shows a distinct query behind them. If it never does, they get folded into the pages they overlap.

What we are testing

Three questions, in order:

1. Do the pages get indexed at all, and how quickly? The domain is not yet verified in Search Console. That is the first thing to do after this deploy, along with submitting the sitemap.
2. Which category earns impressions? Reply examples, guides and post ideas are different kinds of intent. I expect the examples pages to be the surprise, in one direction or the other.
3. Does any of it convert? Every call to action on these pages carries a placement tag, so a trial that starts from a guide is distinguishable from one that starts from the landing page. Free-to-paid will take longer to read because the trial is seven days.

What the first batch taught me

Two things, both about process. The validator caught more in the first hour than I expected: descriptions that ran past 160 characters on almost every page, and a handful of phrases that looked like statistics but were really just confident writing. Both are the kind of thing a human reviewer skims past at page thirty. And writing five audience variants in parallel, with a duplicate check waiting at the end, forced each one to be organised around real situations instead of a shared outline, which is the opposite of how most programmatic content gets made.

What I will not claim

Nothing about rankings, traffic or signups until the numbers exist. The pages went live today. If they rank, I will write that up with the Search Console data. If they do not, I will write that up too, because a validator that keeps content honest is only half the job, and the other half is reporting what happened.

If you want to see the result of all this, the pillar guide on how to grow on X is the page the rest hang off, and what to reply when someone compliments you is a fair sample of the examples pages.