AI Answers and Search Rankings Are Decided in Different Places

Ask a Chatbot About Your Own Company

Doing this once makes the rest of the conversation shorter. Open a chatbot and type the sentence a customer would type. Something like asking who is good at a particular job in a particular city. Read the answer and look for your company in it.

Usually one of three things comes back. Your name is absent, or it is there with a description of a business you stopped running two years ago, or a competitor is named first and you appear at the end of a list.

That screen behaves differently from a page of search results. Search hands over a list and lets the customer pick. Sitting tenth still leaves the chance that somebody scrolls. An AI has already picked, then compressed the choice into a few sentences. If you are not inside those sentences, your position on the underlying list does not do anything for you.

The Foundation Is Shared

Answer engine work gets marketed as an entirely new discipline, and the overlap is bigger than that framing suggests. The material an AI reasons over is still pages published on the open web. A page a search engine cannot read is a page an AI cannot read either. Indexing, titles, page structure, load time: all of it carries over from ordinary SEO. Google’s AI summaries run on top of the same Search index rather than on a separate crawler built for them.

So a company that cannot be found under its own brand name has the order backwards when it asks for answer engine work first. Being readable comes before being quotable. Even when the two are quoted as separate pieces of work, that is the sequence that holds.

Rankings Compete by Page, Answers Take Paragraphs

Search ranks whole pages against each other. An AI lifts a paragraph. That difference changes what is worth touching on the site.

A sentence that answers the question directly has to exist somewhere on the page before anything can lift it. State what the company does in one sentence near the top. Put recurring customer questions in as headings and answer them underneath. Conditions laid out in a table and specifics in a list travel well. Company profiles built out of adjectives do not, because there is no fact in them to carry away. A sentence about what you do, where, and under what conditions beats a sentence about how long you have been in business.

Sentences on other people’s sites count too. Answer engines read what has been written about you elsewhere, which is why a wrong fact often traces back to a directory listing you filled in years ago or an old press release rather than to your own site. When the site has been fixed and the answers have not changed, the leftover sentence is usually still out there somewhere.

Blocked in robots.txt, Then Waiting to Be Quoted

Plenty of owners added AI crawler rules to robots.txt because they did not want their text used for training. Asking later why the chatbots never mention them, the cause may be those few lines.

The confusing part is that the switches are not one switch. Google keeps a separate control for generative AI use, and it sits apart from Search visibility: turning it off does not drop pages from search results, and leaving Search open does not make everything available to every answer product. Crawler names differ by company and so does the scope each one covers, so the first thing to establish is which doors your site currently leaves open.

What has to be decided here is policy, not configuration. Wanting to stay out of training while staying inside the answers is a reasonable preference that the available controls will not always deliver.

Advice circulates that dropping in an llms.txt file is enough to get read. No major engine has published support for it. Adding one does no harm, and it will not get you cited on its own.

Nobody Can Sell You a Guarantee

Ask the same question five times and the wording shifts each time. Change the model and it shifts further. Guaranteed placement in AI answers fails for the same reason a guaranteed number one ranking always did.

There are things that can be measured instead. What the answers say right now, how far they drift across repeated runs, how many statements are factually wrong. Those go into a document, and running the same battery later gives you a comparison. Asking a vendor what they intend to measure usually sorts the field quickly.

The Questions Where Your Name Is Missing

Weple’s AI search visibility work starts by writing the question list. We build questions that fit your industry, put them to the major AI services repeatedly, and collect what comes back along with every statement that is factually wrong. After that we rework the site structure and the wording into a shape that is easier to quote. Your company name and site address are enough for us to send back a sample of what the answers look like today. If you already have a screenshot of an AI describing your business, that screen is where we start.