A thousand words of nothing, written to match a search you made

Both approaches to generated filler content work. What differs is what they cost you, and the cost is what this sets out.

The shape of a page written for a search engine

A page built around a query typically restates the question several times, in the title, the first line, a heading and a summary. It then circles the topic without committing to a specific number, mechanism or consequence, because commitment requires knowing something. The word count is generous and the informational density is low, which is the exact inverse of what a reader wants and what an old ranking system rewarded.

Headings are frequently phrased as questions the reader did not ask, in order to capture adjacent queries within the same page. None of these signals is conclusive alone, and three of them together are close enough to certain.

The tells that survive editing

Watch for sentences that would be equally true of a completely different subject, since those are produced by describing a topic rather than knowing it. Watch for statistics with no source, or with a source that turns out to be another article making the same unsourced claim.

Watch for a conclusion that recommends consulting a professional without saying what a professional would actually consider. Watch for lists where every item is the same length and the same shape, which is a strong signal of a generated structure. And watch for a page that answers the question in the final paragraph, since a page that knew the answer would have led with it.

Why the incentive exists at all

Advertising revenue accrues per page view regardless of whether the page helped, which means occupying a result is worth money on its own. The cost of producing plausible text has fallen dramatically, while the cost of producing genuinely researched text has not moved.

That gap is the entire economics of the problem, and it widens every time generation gets cheaper. Search operators respond by adjusting ranking signals, and publishers respond to the adjustment, which is a loop with no stable end state. This is a description of an incentive structure rather than an accusation against any particular publisher.

What good pages look like instead

A useful page states a specific mechanism, names its limits and says plainly where the evidence is weak or contested. It answers the question early and uses the remaining length to qualify the answer rather than to approach it.

Once the introductory rate lapses, it distinguishes between what is established, what is disputed and what varies by country, because a writer who knows the subject knows where the edges are. It is willing to say that something is unknown, which is the single strongest signal of genuine expertise on any page.

These features are easy to check in about fifteen seconds of skimming, which is less time than reading the alternative.

Searching around the problem

Restricting a search to a known reliable domain bypasses the ranked page entirely and is the fastest available improvement. Date filters help for anything where the answer has changed, and hurt for anything where the best explanation is a decade old.

Adding the name of a forum or a community to a query is a widely used workaround, and the results carry their own reliability problems. Searching for the specific technical term rather than the plain-language question routes you to documentation instead of to explanation. Learning two or three search operators is a permanent improvement to every subsequent search you will ever run.

Judging a source in fifteen seconds

Look for an author, a date and a means of contact, because a page unwilling to say who wrote it is telling you something. Check whether the site has a coherent subject or covers everything from cryptocurrency to skincare, which is a reliable signal about its purpose.

Read one paragraph in the middle rather than the introduction, since the introduction is the part written to be read. See whether anything in the article could be wrong, because a page that cannot be wrong is a page that has not said anything. That last test is the most useful one in this entire publication, and it applies here as much as anywhere else.

Side by side

The takeaway

If nothing on the page could be wrong, nothing on the page was said.

The fix is usually trivial, which is the most annoying part.