The keyword that looks perfect and sells nothing

Every keyword tool answers the same two questions: how many people search this, and how hard is it to rank. Both are useful. Neither tells you whether the people searching would ever buy what you sell.

That gap is where most automated blogs quietly fail. Not on grammar, not on structure, not on word count. They rank, sometimes quite well, for phrases that bring in an audience with no reason to become a customer.

A worked example

Take a phrase like automated SEO reports. On any keyword tool it looks like a gift.

A high cost per click on a low-difficulty keyword is normally the strongest buy signal you can get. Advertisers only bid $8 when the click converts.

Now open the results page. The first page is agency reporting dashboards, white-label PDF builders and Looker Studio template packs. Every single result sells the same thing: a way for an agency to send a branded monthly report to a client.

If you sell content generation, you can write the best page on that list and still lose. Not because the article is weak, but because the person who typed that phrase wanted a reporting tool and you handed them a writing tool.

Why the metrics lied

They did not lie. They answered a different question than the one you asked.

Volume counts how many people type a string. Difficulty estimates how contested the ranking is. Cost per click reflects what the current set of advertisers will pay, and those advertisers are the reporting tools, not you.

The only artefact that encodes intent is the results page itself, because it is the aggregate answer of everyone who clicked something and stayed.

What to check instead

Before writing, pull the live results and ask four questions.

1. What category do the ranking pages belong to? Not the topic, the category. Reporting tools and writing tools can share a topic and share no buyers.
2. Would a person on those pages recognise your product as an alternative? If the honest answer is no, the keyword belongs to someone else.
3. What format is winning? A results page full of comparison tables is telling you it does not want an essay.
4. Is there an AI Overview, and who does it cite? That box is increasingly the answer, and the citation list is the shortlist you are trying to join.

If the first two questions fail, stop. The cost of skipping a keyword is zero. The cost of writing for it is a page that dilutes your topical focus and never converts.

The cheaper mistake

There is a second failure that looks like success. You pick a keyword that matches your audience perfectly, then write the same article a competitor already wrote, with the same sections in the same order.

That page is not wrong, it is redundant. Search engines have no reason to promote a copy over the original, and readers have no reason to remember you.

The brief is where this gets fixed. Pull what actually ranks, list what every result covers, and then find the thing none of them says. That gap is the article. Everything else is filler you have to include for completeness, not the reason anyone will link to you.

Doing this at scale

The check above takes a careful person about ten minutes per keyword. At one article per day that is a standing appointment nobody keeps past week three.

Automating it is not hard, but the order matters:

• Research keywords in batches, not per article. The metrics move slowly and running the same lookups thirty times a month is thirty times the cost for the same numbers.
• Fetch the results page fresh, per article, right before writing. This is the part that cannot be cached, because it is the part that changes.
• Run the intent check before any writing happens. A rejected keyword should cost you one cheap lookup, not a full draft.
• Log every rejection with its reason. A pipeline that silently skips work is indistinguishable from one that is broken.

That last point is the one people skip, and it is the one that makes the system trustworthy. If you cannot audit why a keyword was dropped, you will eventually stop believing the ones it kept.

The uncomfortable part

Doing this properly means publishing fewer articles than you planned. A queue of thirty keywords will not produce thirty articles, because some of those keywords belong to another category and you only find out by looking.

That is the correct outcome. Twenty articles aimed at people who can buy from you will beat thirty aimed at an audience assembled by a difficulty score. The number that matters was never the number of posts.