Google redefined "main content." The query-matrix site just died
Google quietly changed a definition, and the fallout for sites is massive.
Used to be, main content meant the article text. Write 2000 words, job done, page exists. Now Google says it straight: main content is whatever directly helps the page do its job. Text, video, calculator, table, real reviews, a filter — the form doesn't matter. What matters is whether it solves the user's task or not.
Looks like a small wording tweak. In reality, it's a death sentence for an entire class of sites built on keyword matrices.
How sites were built before, and why it stopped working
The classic scheme: grab a keyword list, stamp out a page for every long-tail variant. The text is nearly identical, words shuffled around, the right keywords dropped in. That's not content. That's the same thing repackaged under different URLs.
Google now scores completeness of the solution, not word count against a query. If a calculator page has no working calculator, that's not main content — that's text wrapped around an empty hole. That page won't pass the check, no matter how many expert-sounding words sit on it.
I've said this for a long time: give the user what they need first, get the traffic after. Not the other way around. Under that logic, a keyword matrix is an attempt to take without giving anything. Google sees that now, and it cuts.
Fake authors deserve their own paragraph
In plenty of niches this is almost standard practice: the page features "expert John Smith, 10 years in the industry," and in reality that name was generated along with a stock photo. Google calls this spam outright, if the profile exists only to fake expertise rather than reflect real work behind the content.
If you don't have a staff writer with a real bio, write from the editorial team or the organization. Honestly. No invented John Smith. It's simpler for you, and it doesn't put you at risk under algorithmic review.
What actually needs to change
First, remove anything resting on trust in a nonexistent expert, and replace it with editorial framing or real people with verifiable experience.
Second, pull dynamic facts — prices, dates, specs, terms — from verifiable sources instead of generating plausible-sounding text. This matters most where conditions shift fast: an outdated number isn't a typo, it's a direct deception of the user.
Third, rebuild the architecture. Instead of a hundred near-identical pages for every long-tail query, build one resource that closes the whole task: comparison, calculator, filter, real reviews, data that actually gets updated. Inside one URL, not smeared across a matrix.
Fourth, stop churning out pages for the questions you assume people ask AI search. Same trick, different sauce — content made for a system to be fooled, not for a human to be helped.
And before you scale a new template to hundreds of pages — test it on a limited batch first. Watch indexing, impressions, errors, actual useful user actions. That's cheaper than shipping a thousand pages and cleaning up the drop afterward.
How to measure quality now
Forget word count as a metric. Ask yourself: what does this page solve, what data does that require, what choices does the user actually have, what are the constraints, and is there proof any of it works. If you promised a calculator, it has to calculate. If you promised reviews, they have to be real. A broken tool zeroes out any text around it, no matter how well it's written.
This isn't a one-time cleanup. It's a shift in how you build a resource in the first place. For those ready to rebuild site architecture around the new rules instead of patching symptoms, we help with audits and structural rebuilds in our services.