Generative AI drives the value of consensus content to zero

If you took the rollout of AI search features as permission to mass-produce more checklist SEO content, your traffic is going to disappear. Generative AI drives the value of consensus-restating content to zero. If your content strategy survives contact with ChatGPT, it is because it contains something ChatGPT does not have. You have to stop optimizing for compliance and start optimizing for information gain.

A perfect content score means you reproduced the existing consensus

Compliance SEO is what most people picture when they think about search optimization. You complete a checklist and score well on a test where the answers are already defined. A content-score tool reads the top ten results, lists the subtopics and terms those pages share, and grades your draft against that list. The tool does not care about meaning. You write the page the grader wants and you get a green 92.

The test’s answer key is the set of pages that already exist. If ten competitors all use the same content grader, they all get a 92. The search engine now has eleven identical pages, and it has no reason to prefer yours over the one published three years ago.

Since the answers are defined before you start writing, anyone can score well. A generative AI model will do it in about thirty seconds. Work defined by compliance is imitable by definition. If your entire strategy is summarizing the top ten results, you are basically playing Family Feud. You are guessing what the survey said, and the prize is zero traffic.

Information gain asks what you know that a model does not

Information gain flips the starting question. Instead of asking what the checklist wants, you ask what you can tell the customer that they cannot get by prompting an AI. What do you know that the web does not?

The answer cannot take the form of better prompts or faster execution. Those are easy to imitate. The answer has to be grounded in actual differentiation. You need proprietary data from your transaction records, your support tickets, your measurements, or your internal benchmarks. You need firsthand experience of doing the work and the judgment you formed from real projects. If you sell industrial coffee machines, your support tickets contain the exact sequence of parts that break first. Your internal benchmarks show exactly how many days a machine runs before a specific pump goes kaput. The internet does not know these facts until you publish them.

Google is not indifferent between these two approaches. The company holds a patent on scoring documents by the new information they add beyond what a user has already seen. Its AI optimization guide tells site owners that AI systems look across many sources, that a unique viewpoint stands out, and that a first-hand review beats a summary restating available information. The guide states directly, “Don’t just recycle what others on the internet have already said, or could easily be produced by a generative AI model.”

Compliance is the floor and the floor is now free

“My content grader and checklist exist because they work. Pages that score well do rank.”

That is a valid objection. Covering a topic properly is table stakes, and a page missing the basics will lose. But compliance is the floor. Because a language model reaches that floor in seconds, none of that baseline work is why anyone ranks durably or gets cited by AI assistants. The differentiating margin always came from what sits on top of the floor.

That margin is exactly what a checklist cannot contain. A checklist can only encode answers that already exist. I made the general argument against checklist optimization in an earlier post. When it comes to content specifically, checklist optimization is a bad explanation of how search works. It assumes the search engine wants a perfectly average document. The engine actually wants the document that answers the user’s question most accurately.

If you hire someone to paint your house, you expect them to bring paint. You do not give them a medal for showing up with cans. Covering the basic subtopics of a keyword is bringing the paint. The information gain is the actual labor.

Publishing original claims means risking being wrong in public

Marketers prefer compliance scoring because the green number is visible progress. It requires no access to the business’s own data, and you can defend it in a report. If the page fails to rank, you point to the score and say you did your job. The tool says you got an A. The failure becomes Google’s fault, or the algorithm’s fault, or whatever. The playbook era let marketers avoid hard questions about what the company actually knows because filling templates produced acceptable results.

Information gain is harder to start. It requires talking to people inside the business. You have to extract what they know and publish claims nobody else has published. That means it is possible to be wrong in public.

Problems are inevitable when you publish new ideas. That is how knowledge grows. You put a conjecture out there, subject it to criticism, and correct the errors. A checklist avoids error by avoiding original claims entirely. The era of the playbook is ending. Marketing has to become more like actual marketing again.

You already have the data you need to start

You do not have to invent new science to add information to the web. You just have to mine what the business already has.

Places to find proprietary facts in your business:

• Sales call transcripts
• Customer support tickets
• Internal usage data
• Project benchmarks

These are publishable facts no competitor and no model possesses. If you want to claim a tactic works, run a first-hand test and measure it. A first-hand test always beats a roundup of other people’s claims.

The test for every planned page is whether it adds facts, data, or firsthand material the corpus does not already contain. If the answer is no, the page is a compliance artifact. You are asking a search engine to rank a document that provides zero new value to the person reading it.