Content production: we rejected 12 of 55 topics, and why

Step zero: check whether you already have this page

Before keyword research, before the brief, before anything: does a page on your own site already answer this question? If it does, writing a second one will not lift your ranking. It splits it between two pages, neither of which is complete.

The method is unremarkable. Search the target phrase with the site: operator and see which of your pages Google considers relevant to it. If something comes back covering roughly the same ground, the decision is to improve that page, not to build it a rival.

We keep this in a file: every rejected topic, with its score and the reason it was rejected. The value is not the record. The value is that three months later nobody proposes the same topic again. Keyword cannibalisation is exactly what this step prevents.

Step one: pick the phrase with a real number

Guessing at search volume is the most common failure in this work. For the Iranian market the usual tools either hold no data or return an error, so the figure has to come from a source that genuinely covers the country. finding real Persian search volume sets out that constraint.

And a rule that has repeatedly saved us work: high volume with heavy competition is a trap, not an opportunity.

A phrase with three thousand searches a month whose first page is held by ten specialists means page four forever for a newer site. A phrase with a thousand that nobody has answered honestly is the better buy.

Step two: have something the others do not, before you write

This is the step that takes the most time and returns the most. Before a single word, you should be able to name the sentence only you can write: a number you measured, a mistake you made, or a limitation a competitor cannot afford to admit.

If you have none of those, your article is the eleventh version of something ten people published. In that case reject the topic and go to one where your hand is stronger. Rejecting a topic is not the process failing. It is the process working.

Step three: write it, then cut it

The five-thousand-word median across those 12 pages is not a decision. It is a habit. Length should come from the question: a question with a three-paragraph answer deserves a three-paragraph page. If your article loses nothing when the introduction and the conclusion are deleted, neither was needed.

What we look for in editing: a paragraph that can be removed whole, a sentence that repeats the previous one in different words, and any number without a source. The third does not get deleted, it gets checked. If the source cannot be found, the number goes.

Step four: ship against a checklist, not a feeling

Before publishing, check the things that are expensive to fix later: title and meta description within the right length, a short ASCII URL, every internal link opening, and image alt text that genuinely describes the image. Our on-page checklist carries the fuller list.

One item is specific to Persian and worth stealing for any non-Latin script. Text pasted out of a word processor or a rendered page almost always brings invisible characters with it. Open the draft in a plain editor before publishing. It takes ten seconds and it prevents a pointless difference between this page and the rest of your site.

When not to write content at all

When your selling page does not work yet. If visitors arrive and the service page has no deliverables, no price and no route to contact, more content only wastes more visits. Fix the page that is supposed to sell, then build traffic to it.

And when nobody is there to continue. Three articles in year one and nothing in year two is worse than having no blog, because it tells a visitor the place was abandoned. If you are not confident it will continue, put the effort into the copy on your main pages instead, which does not go stale.

What are the steps?

Check the overlap with existing pages, pick the phrase from real data, find the angle only you have, write, cut, and ship against a checklist. Their order matters more than their number.

What does content production actually mean?

Building a page that answers one specific question and moves the reader to the next step. A page that answers no question is not content. It is text.

How often should we publish?

At whatever rate you can still sustain next year. Two good pieces a month for three years beats ten a month that stop after six.

How long should an article be?

Long enough to answer the question completely. There is no fixed number, and padding to reach a word count is itself a cause of the quality dropping.

If this process is too heavy for your team, outsourcing it is sensible; just make sure whoever writes has access to that same file of rejected topics. And if you do not yet have a page for the content to lead to, building a site where content converts is the step before this one, not after it.

Method: the numbers about our own process come from this site's topic file and cover 43 accepted topics and 12 rejected ones. The competitor numbers come from Google organic results for the Persian query, depth 20, on 16 August 2026; 12 URLs answered 200 and word counts were taken from the served HTML, so they are approximate and include navigation and footers.

Google's own criteria are in the Creating helpful content guide, and the baseline requirements in Search Essentials.