Read but not named: How to become the sentence AI search writes

What does "becoming the sentence" mean in practice?

It means writing the passage the answer is built from. Concretely, for each question you want to be named in:

1. Find the sentence the answer uses today. Read the answer, look at which source sits beside the sentence that matters, and read that source.
2. Write a better version of that sentence for your own page. Same shape, more specific, with your brand as the subject.
3. Put it where the engine reads first. At the top of the page the engine already cites, or at the top of the section that answers that question.
4. Make it true and checkable. A number with its condition, a date, a source. A vague sentence loses to a specific one.

The GEO paper by Aggarwal and colleagues tested this kind of rewriting on generative engines and reported that methods such as adding statistics, quotations and citations to a source could raise its visibility in the generated answers by up to 40%. The lesson that carries over is not a specific trick. It is that what the passage contains decides whether it is reused.

How do you write a sentence an engine will reuse?

Match the question's shape. A "what is" question wants a definition. A "how much" question wants a number with its unit and condition. A "which is best for" question wants a verdict with its reason. A "how do I" question wants the steps in order.

Make your brand the natural subject. "Attensira's Starter plan starts with a 7-day free trial, card required" can only be reused by naming Attensira. "Most tools in this category offer trials" can be reused without naming anyone.

Include the condition. "On Growth and above", "for teams under fifty", "as of October 2026". The condition makes the sentence precise enough to quote, and stops the engine from generalising it wrongly.

Keep it short enough to lift whole. One or two sentences. If it needs a table to make sense, it is not the sentence.

Say the thing nobody else can. An original number from your own data, your own test, your own policy. The roundup can copy your price; it cannot copy what you measured.

Where should the sentence live?

On the page the engine already reads for that question. That is usually not the page you would pick.

Look at the sources of the answers you are missing from. If one of your pages is among them, as our pricing page was, that page is your best bet: the engine already retrieves it, so a better passage there gets a second chance at every answer. If none of your pages is cited, the first job is getting read at all, which is a different problem: crawl access, a page that exists for the question, and pages elsewhere that point to it. See how ChatGPT reads and indexes content.

If the question is a list question and the engine always builds the answer from a roundup, the sentence you need may belong on the roundup. That is off-site work: getting your one line onto the page the engine already trusts. A direct, specific line is also what an editor is most likely to accept.

Which questions are worth winning the sentence on?

Not every question deserves the effort. A sentence is worth writing when three things are true.

The question is one buyers ask close to a decision. "Best X for Y", "X vs Y", "how much does X cost", "does X support Y". A definition question such as "what is GPTBot" is useful for reach, but rarely changes who gets bought.

Your page is already in the sources. If the engine already retrieves your page for the question, you are competing on the sentence, which is the cheapest fight. If it does not, the work is upstream and bigger.

You have something true and specific to say. A number, a policy, a test result, a clear condition. If your honest sentence would be as vague as everyone else's, writing it will not change the answer.

On our own list, the pricing question passes all three: buyers ask it close to a decision, our pricing page is already in most of the sources, and we can state a specific, true answer. "What is GPTBot" passes only the second. That makes the pricing page the first one to rework and the crawler definition a later one.

A quick way to rank your own list: for each question you track, write down how often you are cited and how often you are named. Sort by the gap between the two, largest first, then drop the questions that are not buying questions. The top of that list is where a better sentence pays first.

How do you measure whether you became the sentence?

Track two numbers per question, on each engine, with the number of answers read behind each:

• Named: the answers whose text mentions your brand.
• Cited: the answers that link to your domain.

A wide gap between the two, cited often and named rarely, is the "read but not named" pattern. Our documentation calls it being "known but not read" when it runs the other way: named but not cited. Both are fixable, and they have different fixes. The first needs a better sentence; the second needs a page the engine wants to link.

Then read before and after. Rewrite the passage, ship it, and compare the next weeks against the week before, counting answers. "Named in 0 of 57" becoming "named in 4 of 57" is a result; anything inside sampling noise is not.

For the full method behind counting mentions as a rate, see how to calculate share of voice.

How does Attensira help you become the sentence?

Cited but not named is a column in Attensira, not something you have to spot: for every answer to a question you track, on ChatGPT, Perplexity, Claude and Google AI, with Gemini from Growth, it records whether you were named, whether you were cited, and which pages it leaned on. Where you are cited but not named, it finds the passage on your page the engine read and drafts a replacement, shown beside the exact text it replaces, as a pull request or CMS draft for your approval. Where the deciding page is someone else's, it drafts the pitch to that page's editor.

That is the service of an SEO agency at the price of a tool. The free audit includes one passage rewrite of your own page, written against the question you lose most.