Information gain: why publishing the same advice as everyone makes you invisible

Information gain is how much new information your page adds to a topic beyond what is already published. It is the reason one page gets cited and another, saying the same thing in different words, gets ignored. The internet does not need another "10 tips" post that reshuffles the top ten results. When machines summarize the web, that page has nothing to offer them.

Most local businesses have the content problem backwards. They think they need more posts. They publish another version of advice that already exists in a hundred places and wait for it to work. It does not work, and now there is a mechanical reason why. This is what information gain is, why sameness went from unhelpful to invisible, and how a business with no research team creates content only it could write.

What is information gain, exactly?

It is the additional information a page provides beyond what a reader has already seen elsewhere. Not how long the page is. Not how well it is written. How much of it is new. Matt Diamante, the founder of the SEO agency HeyTony, puts it plainly:

That is more than a slogan. Google was granted a patent in June 2024 called Contextual Estimation of Link Information Gain, which describes scoring a document by the information it adds beyond what a user has already been shown. A patent is not proof that the exact method is running in live ranking, and Google has not confirmed it is. But the direction is clear, and it matches what Google says out loud. Its helpful content guidance tells you to demonstrate first-hand experience and original information, not to repackage what already ranks (Search Engine Journal, 2024).

There is a test that cuts through it. Take your page and mentally delete every sentence a competitor could have written just as easily. Generic definitions, obvious tips, the stuff that appears on every result for the query. What is left is your information gain. If the page still says something specific and true after that cut, you have some. If nothing is left, you published a rehash with your logo on it.

Why does publishing the same advice fail now?

Because AI systems answer by summarizing the consensus, and the consensus is exactly what your rehash is made of. When someone asks ChatGPT, Perplexity, or Google's AI a question, the system reads the sources that already rank, blends the common answer, and cites the pages that added something the blend did not already contain. If your page is the common answer, you are not a source. You are one of ten copies the model averaged and moved past.

That is the mechanical shift. A classic search engine could rank ten near-identical pages and let you click. An answer engine collapses those ten into one paragraph and cites maybe three or four. Being the eleventh version of a common answer is not a weak position anymore. It is no position at all.

There is now research on which additions actually get a page pulled into an AI answer. A Princeton-led study, the first large academic test of optimizing for generative engines, ran nine tactics across 10,000 queries. The moves that lifted visibility most were not stylistic. Adding original statistics made pages 41% more visible in AI-generated responses. Adding direct quotations added 28%. Citing real sources helped too, especially for pages that started lower down (Aggarwal et al., Princeton, KDD 2024). Every one of those is a form of information gain. None of them can be faked by rewording someone else's post.

This is also why AI search is a channel a local business has to treat as real, now. Nearly half of consumers, 45%, already use AI tools to find local businesses, up from just 6% a year earlier (BrightLocal, 2026). If the tools they use only cite the sources that added something, sameness does not just underperform. It removes you from the answer. For more on that shift, see how AI search is changing local visibility.

What actually counts as information gain?

Original signal, not restated consensus. There are really five kinds that a small business can produce, and they all share one trait: a competitor cannot copy them by reading your page and rewriting it.

• Original data from your own work. A real number from a job, an audit, a small experiment. "We replaced 40 furnaces last winter and 3 in 5 were under-sized" is data nobody else has.
• First-hand experience with specifics. What actually happened, in detail. The step people skip. The part that went wrong. The thing you only know because you were on site.
• A genuine point of view. A clear stance, not hedged middle ground. Say what you would not do and why. Most content is too careful to mean anything, which is why it gets skipped.
• A proprietary process or framework. The way you actually do the work, named and laid out. Even a simple checklist is yours if it came from your own reps.
• A local angle a national site cannot match. Real prices in your city, the bylaw that trips people up, the seasonal pattern you see every year. National pages physically cannot write this.

Contrast that with information rehash: summarizing what already ranks, generic listicles, and AI-spun posts that average the first page of results. Rehash reads fine. That is the trap. It passes a quick skim and adds nothing, so it earns nothing. The scorecard below shows the gap on the same five signals.

How does a local business create it without a research team?

You already have the data. You just have not been publishing it. A local service business runs original experiments every week without calling them that. Every job is a data point. Every quote is a real price. Every repeat problem is a first-hand pattern. The raw material for information gain is sitting in your invoices, your job history, and your own head. The work is turning it into pages.

Start with three sources you can mine this week:

• Your last 20 jobs. What did customers actually ask for, what did it actually cost, and what surprised you? "Most people think X, but on the last 20 installs it was actually Y" is a whole article no competitor can write.
• The questions you answer on every call. The thing you explain to every customer is a page. Write the real answer, with your real numbers, not the generic one already on ten other sites.
• A real before and after. One project, documented honestly. What it looked like, what you changed, what happened. This is the format AI systems and buyers both trust most, because it is verifiable.

This is the same work behind our own proof. When we took Magic At My Door from a 52 to a 90 in site health, and E&M Equipment from a 31 to a 90, the write-ups are not generic redesign advice. They are specific: what was broken, what we changed, what moved. That specificity is the information gain. A competitor can read those case studies and still cannot claim the outcomes, because the outcomes are ours. If you want the deeper how-to on turning this into a content plan, see whether your local business should have a blog and local keyword research for small business.

Is AI-written content bad for SEO?

No, but AI-written rehash is. Google does not penalize content for being AI-assisted. It penalizes content that is unhelpful, and most AI content is unhelpful for the same reason most human content is: it adds nothing. Point a model at the top ten results and ask for a summary and you get a fluent, confident rehash with zero information gain. It reads fine and cites nothing, because it contains nothing new.

There is also a trust cost when the seams show. Consumers can increasingly tell, and they react. In a December 2025 survey of 8,000 consumers, 31% said visibly AI-made marketing makes them trust a brand less, while only 7% said it makes them trust it more (Klaviyo and Datalily, 2025). Visible AI in marketing is roughly four times more likely to cost trust than build it (eMarketer, 2026). Publish generic AI output and you can lose the ranking and the reader at the same time.

None of that means do not use AI. It means do not ship what AI gives you by default. Used well, AI is a fast way to structure, edit, and pressure-test writing built on your own facts. Used lazily, it is a rehash machine. The dividing line is whether you fed it something original to work with. Our own take on the practical stack is in the AI tools worth using as a local business owner.

Where does judgment come in?

Judgment is the whole game, because AI can accelerate the writing but it cannot decide what deserves to exist. It does not know which of your jobs made a good story, which number is worth publishing, or which stance is true. It will happily produce ten more pages of consensus at the push of a button. Volume was never the constraint. Deciding what is worth saying is.

This is the part that does not scale, and that is exactly why it is the advantage. Access to AI is table stakes now. Everyone has the same models. What separates a page worth citing from a page worth skipping is the judgment behind it: knowing which detail matters, what to cut, and what only you could say. AI writes faster than people. It does not know what is worth writing.

So the constraint moved. It used to be production, and AI solved that. Now the constraint is taste, the ability to reject the obvious answer and publish the true one. That is not a disadvantage for a small operator. It is the one edge a content farm cannot buy, because it comes from doing the actual work. For the fuller argument on why sameness is losing across search, see is SEO dead in the age of AI and answer engine optimization for local business.

Frequently asked questions

Publish less. Publish the thing only you could write.

You do not need more content. You need pages worth citing. When the machines summarize everything, the reward stops going to whoever published most and starts going to whoever added something new. A rehash is invisible to a system whose whole job is to skip the redundant part. The delta is the only thing left to quote.

So change what you measure. Not posts per month. Information gain per page. Before you publish, run the cut: delete everything a competitor could have written, and see what survives. If the page still says something specific and true, ship it. If it collapses to nothing, you were about to add one more copy to a pile the machines already ignore. Pick the topic only you can cover, put your real numbers on the page, and say the part everyone else was too careful to say. If you want a read on which of your pages are pulling weight and which are just noise, start with the free audit.