Why an index of AI answers must never sell inclusion

An index of AI answers must never sell inclusion because its only value is that it reflects what the AIs actually said. If a brand can pay to be in, or to be higher, then readers cannot tell evidence from advertising, and every number loses its meaning.

What is an index of AI answers for?

Its job is to answer a question that buyers, agencies and brands all have: when people ask an AI about this kind of product, who gets named, and how? To be useful, it has to be an honest record of the answers.

Think of the difference between a ranking built from observation and a directory built from payment. A directory tells you who paid. An index of observations tells you what happened. Mixing the two ruins the second.

What goes wrong when inclusion can be bought?

1. Readers can't trust any entry. If some brands are there because they paid, you cannot tell which. Doubt spreads to all of them.
2. Money decides who is visible. Big budgets buy a place. Small and new brands, who most need an honest signal, get shut out.
3. The measure turns into a product. An index that earns from the brands it measures has a reason to flatter them and to avoid showing bad news.
4. Brands stop looking at the evidence. If the way in is to pay, the incentive to improve the pages and sources that AIs actually use is weaker.

How does Citewake's published method handle this?

It sets rules, and states them.

• Inclusion only through evidence. A brand enters a category's index when an AI names it in an answer, or when a page an answer cited lists it. Nobody can submit, pay for or request inclusion.
• No overrides. Nobody can buy a place, a number or a correction.
• Special treatment for none, including the owner. The method places aipozi, one of Citewake's own products, in the AI search visibility tools category. It is measured by the same code, has no manual adjustment and is always included in the spot check on each edition. The conflict is stated on the methodology page.
• Corrections are added, not written over. An error is fixed by publishing the corrected figure beside the original, with the date, the cause and the evidence. History stays visible.
• Disputes have a limit. A brand can show an extraction error. If the answer or the page snapshot proves it, the figure is corrected. A brand cannot dispute an AI's opinion.

What about gaming the evidence?

If brands cannot pay for inclusion, they might try to manufacture evidence instead, for example by creating pages designed to be cited. The method guards against that in several ways:

• A page counts only if an answer cited it.
• Weighted presence depends on how many answers cited the page.
• Sudden changes driven only by newly cited, rarely cited pages are reviewed before publishing.

No guard is perfect. The aim is that influencing the numbers requires influencing what the AIs actually say, which is the thing worth influencing.

Can a company sell something else around the index?

Often, yes, and it needs care. A business can run a free index and sell tools to improve results, or sell reports to brands. What it must not do is let that commercial relationship touch the index itself. The test is simple. Would the entry or the number be any different if the brand were a customer? If the answer is yes, the index is compromised.

Readers can use the same test on any ranking or "top tools" list. Ask how entries are chosen, whether anyone paid, and whether the publisher sells anything to the brands it lists. Honest publishers say.

What does this mean for AI visibility work itself?

It points to a useful principle: the surest route to being recommended is to be worth recommending, and to be described accurately on the pages AIs read. That is slower than buying a listing, and it is the one that lasts. How to get recommended by ChatGPT and the kinds of pages AI assistants cite most describe the practical side.

It also affects how you read vendor claims. A tool or report that promises a place in AI answers in exchange for payment is selling something no one can honestly deliver. What can be sold is analysis, tracking and advice. What cannot is the answer itself.

Who is hurt most by pay-to-play?

New and small brands. Established names are already visible, so a paid place adds little for them. A young product with real merit has to be found through evidence, and an honest index is one of the few places where that can happen. When a place can be bought, the index stops being a way in for newcomers and becomes one more cost of doing business for everyone else.

How does independence show up in the method?

In small, checkable ways:

• the questions are public, so anyone can rerun them;
• no scored question names a brand;
• every answer is kept, so a number can be traced;
• numbers come with ranges, so differences are not oversold;
• the limits are stated plainly.

Why every AI answer should be kept and margins of error in AI measurement, explained simply cover two of these. The methodology covers the rest, as a method ahead of the first edition in early 2027. To see how assistants describe your site today, you can try the Citewake checker.