In this article
It is frustrating to spend a working day on an article only to discover that it says the exact same thing as the top five search results. When AI has made it almost free to produce decent, average copy, writing well is no longer enough. If your content is to have a clear reason to be chosen in 2026, it must add something the other answers lack: proprietary data, documented experience, or an angle that helps the reader make a better choice.
In part 1 on how AI models search, you will find the background on how content is retrieved as potential sources.
What does information gain mean here?
The term has a precise definition in information theory. In this article, I use it as a practical content concept: What does your page add beyond what already exists?
A new phrasing is not necessarily new information. You can write an entirely original piece on "five benefits of SEO" without giving the reader a single new insight.
That is precisely why the concept has become so relevant again. When text generation became virtually free, the web was flooded with articles that are accurate, tidy, and completely interchangeable. Both Google and AI services are left with the same problem as the reader: how to separate what genuinely adds value from what merely repeats?
In May 2025, Google recommended prioritising unique content that is not merely commoditised filler. The guidance explicitly covers AI Overviews and AI Mode as well.
That does not mean Google has published an information gain score you can optimise against. Nor is it proof that all AI models select sources using the same blueprint.
The practical test is simpler: If this page disappeared, what knowledge would the reader lose?
Your competitors can copy your text. They cannot copy your experience.
Proprietary data: show how you reached the answer
You do not need a nationwide survey. A focused dataset from your own operations can answer a question better than generic industry advice.
A tradesperson can examine which factors most often delay a project. An online shop can compare return reasons. A consultant can categorise recurring issues found in audited websites.
Yet numbers need context:
- What did you investigate?
- How large was the sample size?
- Which time period does it cover?
- How did you define the categories?
- What can the data not tell us?
Imagine reviewing 30 support tickets. That is a hypothetical example, not an industry survey. The findings might be useful to your clients, but you cannot simply claim they represent all businesses.
Feel free to publish an anonymised table or an explanation of your methodology. Do not publish personal data or confidential client material just to appear more thorough.
Experience: explain the reasoning, not just the result
"I have extensive experience" gives little to work with. A specific description of what you considered is far more useful.
In consultancy work, it is often the trade-off itself that is missing from generic articles. What should you prioritise first? When is the standard advice wrong? Which prerequisites need to be in place?
A useful account of experience follows this sequence:
- The situation and the constraints.
- The alternatives you evaluated.
- Why you chose the route you did.
- What you observed afterwards.
- What you still do not know.
In part 5, I described a specific checkpoint from client setups: the order of channel rules in GA4. It is a minor practical detail, but details like that make advice actionable.
You will find the example in part 5 on tracking AI traffic in GA4.
An experience does not have to end in an impressive percentage increase. A failed attempt can be valuable if you explain what others should look into before trying the same thing.
Fresh angles: find the question standard answers overlook
A fresh angle is not about being contrarian for attention's sake. It should make a specific problem easier to understand.
"How much does the solution cost?" is broad. "When does the cheapest solution become more expensive due to maintenance?" opens the door to a useful comparison.
You can add value by demonstrating:
- Who common advice is not suitable for.
- What changes when working with small budgets.
- Which local conditions influence the choice.
- How two alternatives perform under identical conditions.
In part 4, I drew on my own experience with PromptWatch: I liked the tool's interface, but grew sceptical of what the numbers actually showed. What proved useful was not another feature list, but the distinction between a measured test sample and actual visibility.
You will find that assessment in part 4 on AI visibility measurement tools.
Make your contribution easy to cite
Originality does little good if the reader has to search for the conclusion.
Lead with a direct answer. Place documentation close to the claim. Use descriptive subheadings, and include dates whenever information might change.
Google's documentation outlines how AI features can explore multiple sub-questions across several searches. That is a compelling reason to answer specific questions thoroughly, rather than hiding everything inside a single generic overview.
That does not mean every paragraph needs to be an isolated "AI snippet". Context still matters. Both a model and a human reader need to understand the finding as well as the caveats.
The page must also be crawlable and indexable. Google specifies no extra technical requirements for these AI features beyond standard search criteria. An original investigation behind a blocked page does not solve the accessibility issue.
Start with a single existing article
Pick a page that already answers an important customer question. Compare it against other solid answers, and highlight everything that merely repeats them.
Then add one distinct contribution: a documented experience, a small study, or a comparison with transparent assumptions.
Note the change and the date. Track relevant queries, enquiries, and referral traffic over time. Treat any AI checks as spot checks, not as definitive proof of cause and effect.
The goal is not to make the copy artificially different. It is to give readers a reason to choose your specific answer, even after they have already skimmed a generic summary.
