Buyers now ask ChatGPT and Perplexity which companies to call. Most of the leaders I talk to know this, and they still don’t know what to do about it.
The tools are new, the advice online contradicts itself, and nobody wants to rebuild a website on a guess. Here is what we changed for one client, what we saw, what we can’t claim yet, and where I’d start.
The short answer
Nobody outside these companies knows exactly how ChatGPT or Perplexity picks what to recommend. What you can control is whether good answers about you exist and can be found.
That comes down to four things.
Know the questions your buyers ask. Publish clear, useful answers to them. Keep your facts the same everywhere they appear. Make sure the tools can reach your pages.
None of this guarantees a mention, but each step gives the tools better material to work with.
What we changed for one client
Earlier this year we worked with a client that wanted more of its best-fit buyers to find it online.
We started by asking ChatGPT, Perplexity and Google the questions those buyers ask. For broad questions, the client came up. For the most specific, highest-intent question, the one it was best placed to answer, it didn’t. The answers named directories and much larger competitors.
We also found that the client’s own facts didn’t match. One page said one thing, another said something else, and outside listings repeated older numbers. We can’t see how each tool weighs that, but mixed facts leave a tool without one clear answer to repeat.
So we did four things:
- Agreed on the facts. We built one list of official figures, had the client confirm it, and checked their key pages and main directory listings against it, with a fix for every mismatch.
- Answered the real questions. We wrote articles and short FAQ answers around the exact questions buyers ask, in their words.
- Made key pages clearer. We rewrote the titles and descriptions on nine important pages and prepared structured data for them, so each page states plainly what it covers.
- Measured what we could. We tracked visits from AI assistants in the client’s analytics.
What we saw, and what we can’t claim yet
The client’s Google Analytics reports an AI Assistant channel. From June through September it recorded 328 sessions from AI assistants. About 94% came from ChatGPT, and the rest came from Gemini, Claude and Copilot.
None came from Perplexity in that window. Monthly sessions rose from 47 in June to 119 in September.
It is still early. The account shows no AI Assistant sessions before June, and we can’t tell whether that was a true zero or a change in how the traffic was labeled.
Some of our changes went live during that window, so we can’t say how much of the traffic they caused. The next steps are asking that high-intent question again to see whether the answer has changed, and tying these visits to inquiries.
Where teams get stuck
Without a clear first step, many teams wait, or they buy a tool that reports a visibility score. A score earns its keep when it leads you to fix something.
The one thing to start this month
Choose five important buyer questions and publish clear, useful answers to them.
Ask your sales team which questions come up before a deal. Pick the five that matter most. Answer each one plainly, with the facts right, on a page of your site.
Short video is the most efficient way I know to get those answers out of an expert’s head. Say the question out loud, then answer it in a minute or two, the way you would on a call. We do this with a client now, one question per video.
One recording gives you three things:
- A video you can post with the question as the title.
- A written answer for your website, edited from the recording so it is useful on its own.
- Material for follow-up content, such as a post, an email or a reply your sales team can send.
That’s how we work in general: the expert brings the knowledge and judgment, and a repeatable system turns it into pages, posts and emails.
Before you publish, ask whoever runs your website to confirm that these pages aren’t blocked by site or firewall settings. ChatGPT search uses a crawler called OAI-SearchBot, and Perplexity uses PerplexityBot. Both companies control those search crawlers separately from any crawling used to train AI models, so you can allow one without the other.
How to tell whether it’s working
Save the same five questions and check them every 30 days. That gives you a steady rhythm for spotting changes.
Each time, write down:
- The tool and the date.
- Whether your company was mentioned or recommended.
- Which sources the answer cited.
- Visits from AI assistants in your analytics, and any inquiries that came from them.
Answers change from one try to the next, so a single response is a weak verdict. Look at the pattern over a few months.
Frequently asked questions
Does structured data get you recommended by ChatGPT or Perplexity?
Nobody has shown that it does on its own. It can help a page state clearly what it covers. For Google’s AI features specifically, Google says there is no special schema you need to add.
Do I have to let AI companies train on my content to show up?
No. OpenAI and Perplexity both separate their search crawlers from model training. Your website person can allow the search crawlers and make a separate decision about training.
Comparable reads
- Can AI content convert? Why it’s the wrong question to ask. Why trust and knowing the buyer still close the deal.
- More noise, same buyer: why B2B buyers keep picking the safe choice. What makes a buyer feel safe choosing you.
- Signal before scale: know what works before you do more. Why making more is the wrong first move.
Want a second set of eyes on what AI tools say about you?
Book a strategy conversation with me. Bring the five questions your buyers ask most, and we’ll look together at who gets named.
One Response