You can check whether AI recommends your business in one week, without an enterprise tracking platform. Ulta Beauty treats agents as voices it cannot control and can only inform, while Bolt is tracking how often chatbots name it so new content can win more answers. Searchable’s argument, reported by PPC Land, is that a brand left out of those answers can lose customers to rivals without seeing it happen.
Updated November 8, 2026
Key Takeaways
- Ulta cannot control agent answers; it can only shape what recommenders know.
- Log buyer prompts monthly; one test is not proof.
- Allow search crawlers, then inspect five revenue pages.
- Track qualified demand, not chatbot mentions alone.
Why do AI answers change who gets found?
Thinkpeak’s GEO guide says generative engine optimization is SEO plus three jobs: let AI search crawlers in, publish pages worth quoting, and measure citations. It cites a Pew Research Center study of 900 US adults. Only 1% of visits included a click on a link inside an AI summary.
Alphabet said in its Q2 2026 earnings call that AI Mode has surpassed one billion monthly active users. OpenAI says ChatGPT is used by more than 1.2 billion people every week. Google says clicks from pages with AI Overviews are higher quality, with visitors more likely to spend more time on the site.
| Pew finding | With AI summary | Without AI summary |
|---|---|---|
| Clicked a traditional result | 8% of visits | 15% of visits |
| Ended the browsing session | 26% of pages | 16% of pages |
How do you check if AI recommends your business?
Kelly Mahoney, chief marketing officer at Ulta Beauty, said at the Status Summit, “We say we want to be the best marketers to our consumer, but we also need to be the best marketers to the agent,” according to MarketScale. The brand does not write the recommendation. It can only shape what the recommender knows.
Searchable’s November 6, 2026 announcement says Bolt will track how often chatbots name it on buyer questions, such as an airport ride, across the UK, Germany, Poland, South Africa and Thailand. The team will use those findings to see where new content can push ahead of rival apps’ AI visibility. PPC Land notes the release does not mention paid placements inside ChatGPT, a price, or a baseline for how often Bolt is named today.
Entrepreneur turns the same idea into a spreadsheet. A recommendation can change with the question, location, timing and context, and different tools generate different responses. Treat the log as directional research. One test is not proof that a business is universally invisible.
- Write five to 10 prompts from sales calls, emails and proposals: a category question, a problem question, a comparison and one that names your market.
- Keep branded prompts separate from discovery prompts.
- Test the wording in more than one AI search experience. Log prompt, platform, date, cited source, competitor, opportunity and whether your site is cited.
- Repeat the same set monthly and look for gaps that persist.
What should you fix before adding content?
Inspect five revenue URLs before you add pages. Use Google Search Console to check indexing, and look for a noindex directive, a broken canonical tag, restrictive robots rules, weak internal links and text that is not readable. Assign each issue to a named owner with a completion date.
A page must be indexed and snippet-eligible before it can appear as a supporting link in AI features. No schema package guarantees a mention.
Allow OAI-SearchBot, Claude-SearchBot and PerplexityBot. Disallow GPTBot and ClaudeBot. OpenAI’s documentation, cited by Thinkpeak, says sites opted out of OAI-SearchBot “will not be shown in ChatGPT search answers,” and each setting is independent.
Thinkpeak’s check is curl -A “OAI-SearchBot” -I against the site, confirming a 200, because a plugin can block the agent before robots.txt is read. OpenAI notes changes can take about 24 hours. Remove nosnippet or a restrictive max-snippet on key pages.
Rewrite the first 100 words so they name the audience, problem, service and outcome, under one H1. On each of five revenue pages, answer one pre-quote question in the first sentence of an H2. Explain price variables rather than inventing a price.
Replace “industry-leading” with evidence a buyer can check. Thinkpeak’s summary of a KDD 2024 paper found keyword stuffing scored below the unedited baseline. Check that your business name is consistent in reviews and directories, match it on a marketing-aligned website, and do not use AI-generated articles as a stand-in for proof.
How do you tell a mention from demand?
“I don’t think there’s a playbook, so it’s really stepping through this. But you can’t wait. It’s here,” Mahoney said. She also said “all of this is measurable.”
MarketScale’s suggested test is to pick one problem, such as acne-prone skin, and ask whether owned content, associates and creators would produce the answer an AI agent judges best. Bolt’s release gives no baseline for how often it is named and does not mention bookings. Measure qualified demand, not just mentions. Repeat the unbranded prompts monthly.
Frequently asked questions
Does generative engine optimization replace SEO?
No. Thinkpeak describes it as SEO plus three jobs: let AI search crawlers in, publish pages worth quoting, and measure citations. Unindexed pages are rarely retrieved.
Do small teams need a paid visibility platform?
Not to start. Entrepreneur’s method is a spreadsheet of prompts, platforms, dates, cited sources and competitors. Bolt’s Searchable deal states no price and no baseline.
What if you block OAI-SearchBot?
Thinkpeak cites OpenAI’s documentation: sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. Training-crawler settings are separate.
How should you judge whether recommendations improved?
Repeat the same unbranded prompts monthly and measure qualified demand, not mentions alone. Mahoney said she does not think there is a playbook, but that the work is measurable.
