Hurry up!! your website’s new guest is already at the table.

Hey there!

Think of your website as a dinner party. Your guests used to be people. Now there’s a new one at the table: an AI agent. It doesn’t eat, it reads very fast, and afterwards it tells your customers what it thought of you.

So here’s what’s on the menu to make sure it leaves with the right idea.

- Three questions we’re hearing from marketers in Q4
- Two playbooks worth bookmarking
- One thing worth sharing

Three questions we’re hearing from marketers in Q4

✦  Customers don’t choose the brand they discover first. So what actually decides?

In our experience, it’s the brand that feels most relevant at the moment of decision. People now meet you through AI answers, search, social, reviews and communities, so being found gets you considered, not chosen. What tips it is usually a page that picks up where the ad or answer left off, says plainly who it’s for and what it costs, and shows proof. A quick test: land on your page as the person your last ad sent, and ask whether it feels like the same conversation.

✦ Has anyone on your team checked whether ChatGPT can even open your pricing page?

Probably not, and it takes about ten minutes. Agents tend to read the raw text of a page, often skip scripts, can’t always see images and may stop early, so open your pricing page the way they would.

Check that your site isn’t blocking the bots agents use, that the price and plan details are real text instead of a size picker, an image or a ZIP code widget, that rates, fees, eligibility and refund terms are spelled out in plain words with dates, and that the answer sits at the top in a few words. If an AI can’t read a fact, it fills the gap from somewhere else.

✦ A person wants to feel it. An agent wants to verify it. Which one is your page written for?

Most pages are written for the feeling, which is right for people. But an agent reading the same page is looking for facts it can confirm. For example: a pitch like “lose weight faster” works on a person, while an agent needs the actual results behind it.
So keep the tagline and add a supporting layer of specifics, proof and terms. Healthcare brands know this pattern well: a headline for people, with evidence and disclaimers underneath.

Two playbooks worth bookmarking

Playbook #1

Find the rung you’re stuck on (about 30 minutes)

If the pricing page check turned something up, this shows you exactly where. An agent has to get through five steps before it can recommend you, and each one depends on the last. For your three most important pages, ask in order:

1. Reach: can an agent fetch the page?

2. Read: are the facts in plain text, not scripts, images or PDFs?

3. Understand: does it say what this is, who it’s for and what it costs?

4. Evaluate: is there evidence or a comparison that shows why it’s the better choice?

5. Act: is there enough detail to shortlist or buy?

Fix the first rung where a page fails.

Insight #2

A ten-question scorecard (about an hour a week)

You can’t improve what you only check once. Write down the ten questions your buyers actually ask, then run the same ten in ChatGPT, Gemini, Claude and Perplexity on the same day each week.

For every answer, note whether you appeared, whether it was right, who else appeared and which of your pages were linked. Change one thing, then rerun next week. AI answers vary from run to run, so look for patterns over a few weeks, not a single result.

One thing worth sharing

Our CEO Ankur sat down with Greg Kihlström on The Agile Brand podcast to talk about a gap most websites have: a page can make perfect sense to a person and still leave an AI agent guessing.

Every agent also arrives for a different reason, so the right version of your page depends on which agent is visiting and what it’s looking for. The conversation is about building one website for both audiences, then testing and learning as you go.