How does AI choose which brand to recommend?
How does AI choose which brand to recommend? It starts from what it learned from public text before a cut-off date, where brands that are written about often, in many places, stand out. For anything newer it searches the web, often turning one question into several searches, and pulls out the passages that seem to answer it. Then it compares what different sources say and writes one answer. If your brand wasn't in what it read, you're not ranked lower. You're simply not there.
What does AI already know about your brand?
Before your customer asks anything, AI has already learned from a very large amount of public text: articles, reviews, forums and product pages. A brand that many different places write about, again and again, stands out in that memory. A brand that is rarely mentioned stays weak in it.
For your brand, this means your own website is only one of many sources. What reviews, articles and forum posts say about you also shapes what AI knows. If your competitors are written about in more places, they stand out more. And if AI read little about you, or only old information, that is the picture of you it repeats.
This memory also stops at a date, called the cut-off date. A new menu, a renovated room or a product you launched after that date is not in it. To know about anything newer, AI has to search.
How AI searches the web
When AI searches, one question from a customer often becomes several searches, each from a different angle. It does not read your pages from top to bottom, as a person would. It takes out the passages, the short parts of text, that seem to answer the question.
So each important passage on your site should make sense on its own. Say who you are, what you offer and where, in plain words.
Google ranking still matters here. In its guide "AI features and your website", Google says its AI answers are built on its normal search ranking. A page that ranks well is more likely to be read. The search work that helps your pages rank is still worth doing.
Pages AI cannot read
Some websites only show their text after code runs in the browser. People see a normal page, because their browser runs that code. Many AI crawlers, the programs that collect web pages for AI, do not run it, so they never see that text.
A website can also block AI crawlers. If yours does, the AI behind those crawlers leaves your site out of its answers. If you block them, make sure it is a choice you made on purpose.
What makes AI use a source
AI looks for the best fit for the question. It compares you with your competitors on price, on reviews and on what each of you is good at. It also uses what the person has told it, such as their budget or who they are buying for.
So say clearly who you are a good fit for. Clear, specific facts are easier for AI to use than general praise. "Family rooms with a kitchen, next to the ski lift" gives AI something to compare. "Something for everyone" does not.
One study (Aggarwal et al., KDD 2024, listed in the sources below) found that pages showed up more often in AI answers after they added sources, quotes and figures.
What to do first
1. Ask AI the questions your customers ask. Read what it says about your brand today, and which competitors it names.
2. Ask the person who runs your website if your key pages show their text without code running in the browser.
3. Check that your site does not block AI crawlers, unless you decided to.
4. Write clear, specific facts on your key pages. Add sources, quotes and figures where you have them.
5. Keep doing the search work that helps your pages rank in Google.
Start with the first step. Ask AI what your customers would ask, and see whether your brand is in the answer.
Transcript
Your question and what the AI already knows (0:00)
You ask an AI which brand to choose. That orange light is your question. Let's follow it. Long before you asked, the AI learned from a very large amount of text from the public internet. Articles, reviews, forums, product pages. It didn't keep copies of the pages. It learned patterns. Which words and names tend to appear together. Names written about often, in many places, became bright points in that memory. Those rarely mentioned stayed faint. One of them could be your brand. And the memory stops at a date. On its own, it knows nothing newer.
It goes looking (0:39)
So it goes looking. Your one question often becomes several searches, each from a different angle. It doesn't read pages the way you do. It pulls out the passages that seem to answer the question.
Some pages stay dark (0:51)
Some pages stay dark. If your content only appears after code runs in the browser, many AI crawlers never see it. If your site blocks them, it is left out of their answers.
How it weighs what it found (1:03)
Then it weighs what it found. It is looking for the best fit for what you asked. It compares what different sources say. Price, reviews, what each choice is good at, and anything you told it about yourself. Google says its AI answers are built on its normal search ranking. So pages that do well in search are more likely to be read. Clear, specific facts are easier to use. In one study, pages that added sources, quotes, and figures showed up more often in AI answers.
One answer: who makes the list (1:33)
Then it writes one answer. Not every brand makes the list. If your brand wasn't in what it read, you're not ranked lower, you're simply not there. And if what it read about you is thin or out of date, that's the version of you it repeats.
See where the light reaches your brand (1:48)
Every day, Avisible asks AI about your market, the way real people ask. We show where the light reaches your brand, where it doesn't, and what to do first. You also see who AI recommends instead of you, which sources it uses and whether your changes are working. Avisible. Try it free for seven days.
Questions this film answers
How does AI decide which brand to recommend?
It combines what it learned before its cut-off date with what it finds in a live search. It compares what different sources say about price, reviews and what each choice is good at, then writes one answer.
Why doesn't AI mention my brand?
If your brand wasn't in what the AI read, it isn't ranked lower: it is simply not there. Brands that are rarely written about stay faint in its memory, and pages it cannot read are left out.
Can AI crawlers read my website?
Many AI crawlers never see content that only appears after code runs in the browser. If your site blocks them, it is left out of their answers.
Does ranking in Google still matter for AI answers?
Google says its AI answers are built on its normal search ranking, so pages that do well in search are more likely to be read.
What makes a page more likely to be used in an AI answer?
Clear, specific facts are easier to use. In one study, pages that added sources, quotes and figures showed up more often in AI answers.
How can I see what AI says about my brand?
Avisible asks AI about your market every day, the way real people ask. It shows where your brand appears, who AI recommends instead of you, which sources it uses and whether your changes are working. Try it free for seven days at avisible.co.
Sources for the facts in the film
- Carlini et al. 2022, "Quantifying Memorization Across Neural Language Models"
- Kandpal et al., ICML 2023, "Large Language Models Struggle to Learn Long-Tail Knowledge"
- Google Search Central, "AI features and your website"
- Vercel and MERJ, "The rise of the AI crawler"
- OpenAI, "Overview of OpenAI Crawlers"
- Aggarwal et al., KDD 2024, "GEO: Generative Engine Optimization"
Avisible asks AI about your market every day, the way real people ask. It shows where your brand appears, who AI recommends instead of you, which sources it uses and whether your changes are working.