Most SEO advice is about to stop working. Not because Google changed the rules, but because the answer moved. When someone asks an AI a question now, they get one answer, not ten blue links, and that answer is stitched together from whoever said something specific enough to be worth quoting.
Commodity content, the kind that restates what everyone already wrote, has nothing to quote. It was already the weakest content on the web. AI search just made that weakness fatal. This is the piece I wish someone had handed me four years ago, when I thought SEO was one simple thing: getting a page to rank. The reality has so much more nuance than that, and most of it now lives in a single question. When an AI answers, whose work does it name?
Non-commodity content is writing only you could have produced: a specific story, a hard number, or a named framework drawn from your own work, rather than a restatement of advice that already exists everywhere. Because AI search assembles one answer from many sources, it cites the page with the irreplaceable detail, so non-commodity content is the most reliable way to get named in an AI answer.
Commodity content is dead weight now
Here is what changed. In the ten-blue-links world, you could rank a page by being thorough enough. Cover the topic, hit the word count, add a table of contents, and you had a shot, even if every sentence existed somewhere else already. The engine ranked pages, and a competent page could win.
AI search does not rank pages against each other. It reads everyone at once and writes a single answer. So the question is no longer "is my page good enough to beat the others." It is "did my page contain the one sentence the model needed, the one it could not get anywhere else." That is a completely different bar. Thorough is not enough anymore. Specific is the whole game.
This is not a forecast. It is already here. By late 2025, depending on whose tracker you believe, an AI Overview was sitting on top of somewhere between a quarter and a half of all Google searches, and the rate climbs the more specific the question gets. For the money-and-health questions where people most want a trustworthy answer, it is past sixty percent. The harder and more specific the question, the more likely a machine answers it before anyone clicks a thing.
I call the content that clears that bar non-commodity content: the article that says something only you could say. It is less a tactic than a test. Am I writing what is already on the internet, or what is only in my head?
The method: demand, the gap, and your real work
A non-commodity article lives at the intersection of three things. Miss any one and the piece either gets no traffic, gets buried under stronger sites, or has nothing worth citing. I think of it as three bowls you are sorting into.
One, demand. What are people actually searching, and now, actually asking an AI? Real volume, in their real words. I pull this with keyword data before I write a single line, because hope is not a search query.
Two, the gap. Look at who is already ranking. If the first page is commodity listicles from sites you cannot outrank, walk away. But often it is wide open. One thing I keep seeing: Google is force-elevating Reddit threads to the top of question-shaped searches, which is Google quietly admitting that lived experience beats brand listicles. That gap is the opening, and it is not a hunch anymore. Across roughly 30 million AI-answer citations, Reddit now ranks as the single most-cited domain on the web, ahead of YouTube, LinkedIn, and even Wikipedia. The engines are not reaching for the polished brand page. They are reaching for the person who actually did the thing.
Three, your real work. The specific story, the hard number, the named framework that is yours and nobody else's. This is the only bowl you cannot fake, and it is the one that gets you cited.
If you cannot say something only you could say, do not write the article. Go find the thing only you could say first.
One more rule that comes straight from AI search. Almost every commercial query now shows an AI Overview, which means ranking number one is not the finish line anymore. The finish line is getting quoted. So I write a tight, fact-dense opening passage into every article, the kind an AI can lift word for word, with the definition and the hard number sitting right at the top where the model cannot miss them. Make the quotable thing easy to find and you make yourself easy to cite.
What this looks like with receipts
I worked with a founder-fundraising specialist who had the thing most content does not: receipts. Real campaigns, real outcomes. The commodity version of her content would have been "how to find investors," a phrase a thousand sites already own and an AI can answer without her.
So we did not write that. We built around her actual framework, which she describes as matching investors by mandate before sending a single email: industry, stage, raise type, and check size, all lined up first, so the outreach is not cold, it is exactly what the investor was already looking for. Then we backed it with her own numbers.
The receipts behind the framework: 3x ROI on a single raise in under a month, a 5.47% conversion across roughly 3,700 investor contacts, and more than $5M raised across the campaigns it came from.
Here is why those numbers matter more than anything else on the page. An AI can paraphrase generic fundraising advice from anywhere on earth. It cannot invent her 5.47%. That number only exists in her work, so when the model wants to say something concrete and credible about finding investors, her page is the one it has to cite. Everything vague about fundraising is replaceable. Her 5.47% is not. Be exact about something true and you stop being a page the engine can route around.
The other half: who said it
Specificity gets you cited. But in the highest-trust topics, your money, your health, your legal standing, there is a second question the engine asks, and it is just as important: who is saying this, and can I trust them?
I audited a tax attorney once, in maybe the hardest niche there is for trust. His credentials were genuinely elite. A former IRS agent who earned his CPA while inside the agency, a law degree, admitted to three different federal tax courts. That is exactly the authority Google and the AI engines want to see on a page about fighting the IRS. And his site was throwing almost all of it away.
His articles, a whole book's worth, carried no credentialed byline and no author schema tying them to the former-IRS-agent who actually wrote them. The trust signal was sitting right there, invisible to the machine. The fix was not more content. It was making the authority machine-readable: a real author entity, a credentialed byline on every piece, his bio and bar admissions structured so an engine could read them and believe him.
In AI search, who said it matters as much as what was said. If your expertise is not machine-readable, the engine cannot give you credit for it.
Write the thing only you could write
Strip away the jargon and non-commodity content is just honesty, said specifically enough to be useful. Your real work, put as plainly as you can stand to put it, made legible to the machines now deciding who gets named. That is the whole job. It is slower than spinning up another listicle, and when the search engine is an AI, it is the only kind of writing that still earns a mention.
I am not the loudest voice in this space. But I am in it every day, on real sites, watching what actually moves. And the thing that moves is always the same: the sentence nobody else could write.
