When AI agents do the buying, who sees your ad?

Think about the last time you asked an AI assistant for a recommendation. Maybe it was a laptop, a dentist, a project management tool. You got a short answer with three or four names and reasons. You didn’t see a single ad.

Now think about where your marketing budget goes.

The short answer: AI assistants like ChatGPT, Claude, Gemini and Perplexity increasingly sit between your business and your buyer. They read the web, compare options and, more and more often, complete the purchase. Their recommendations are mostly earned, not bought, so an interruptive ad has no one to interrupt. To be chosen, your business has to be easy for an AI to find, understand and trust: clear facts, consistent information, real reviews and content that answers the questions buyers ask.

This is part of our series on why ads are fading and distribution is replacing them.

The buyer has a new assistant

Search used to work like this: a person typed a query, scanned a page of links and ads, and clicked. Advertisers paid to be on that page.

Now a growing share of research happens in a conversation. Someone describes what they need, and an assistant does the scanning on their behalf.

The effect on clicks is already visible. Pew Research found Google users clicked a traditional result in 8% of visits when an AI summary appeared, compared with 15% when it didn’t, and they very rarely clicked the sources the summary cited. In early 2024 Gartner predicted a 25% drop in traditional search volume by 2026 as people turned to chatbots and agents.

With an AI summary on the page, clicks on a traditional result nearly halve: 8% of visits, down from 15%. Source: Pew Research Center, July 2025.

Fewer people on the results page means fewer people seeing results-page ads.

Agents are starting to check out, too

It’s not just research. The infrastructure for AI agents to buy is arriving:

In September 2025 OpenAI and Stripe launched Instant Checkout in ChatGPT and the Agentic Commerce Protocol, an open standard for letting AI agents complete purchases with merchants. OpenAI says the product results ChatGPT shows are not sponsored and are ranked on relevance.
The same month Google announced the Agent Payments Protocol (AP2), an open protocol for agent-led payments, backed by payments and commerce companies.

Put those together and you get a buyer journey with no feed to scroll and no banner to notice. An agent gets a brief (“a waterproof running jacket under $150, delivered by Friday”), compares what it can find, and acts.

You can’t bid your way into that answer. You have to deserve it.

What AI assistants look for

No AI company publishes a complete ranking formula, and they change often. But the inputs are visible in the answers themselves, and the principles are stable. An assistant can only recommend what it can find, understand and trust.

Can it find you?

Your site is crawlable. If your pages are blocked, hidden behind scripts that don’t render, or locked behind logins, assistants can’t read them. Check your robots.txt and your CDN settings. Cloudflare, for example, began blocking AI crawlers by default for new sites in July 2025, so some businesses are invisible without realising it.
You exist beyond your own site. Listings, directories, marketplaces, review sites, partner pages and press mentions are all places an assistant may look.

Can it understand you?

Plain facts, stated plainly. What you sell, who it’s for, where you operate, what it costs, how long it takes. If a human has to email you to find out, an agent will move on to someone who published it.
Pages that answer real questions. “How much does a kitchen refit cost in Leeds?” is a question. A page that answers it honestly is far more useful to an assistant than one that says “contact us for a quote”.
Structured data. Markup like schema.org for your organisation, products, prices, opening hours and reviews makes facts easy to extract.
Consistency. If your hours, prices or address differ between your site, your Google profile and a directory, the assistant has to guess. Guessing lowers confidence.

Can it trust you?

Real reviews and mentions from customers and other businesses.
Evidence over adjectives. “Award-winning” and “best in class” tell an assistant nothing. Specific detail does.
Content with a point of view and genuine expertise, rather than thin pages written to catch keywords.

None of this is a trick. It’s what made a business easy to recommend by word of mouth, written down where software can read it.

A practical checklist

If you want to know how your business looks to an AI assistant, start here:

Ask them. Put the questions your buyers ask into ChatGPT, Claude, Gemini and Perplexity. Are you mentioned? Is the information right? Who’s recommended instead, and why?
Check access. Make sure your robots.txt, firewall and CDN aren’t blocking the crawlers you want to be read by.
Publish the facts. Prices or price ranges, service areas, delivery times, policies, what’s included and what isn’t.
Write the answers. One clear page per important buyer question, with the answer near the top.
Add structured data for your organisation, products and services.
Fix inconsistencies across your site, listings and profiles.
Earn mentions. Partnerships, guest content, community involvement and happy customers who leave reviews.

Notice that almost nothing on this list is an ad. It’s distribution: owned content, earned trust and a clear public footprint.

Will there be ads in AI assistants?

Probably, in some form. Some AI products are already experimenting with sponsored placements, and it would be naive to assume answers will stay ad-free forever.

But even if paid slots arrive, the underlying dynamic holds. An assistant that recommends bad products to paying advertisers stops being trusted, and people stop using it. The organic recommendation, earned on merit, is what makes the product valuable. That’s where you want to be.

The lazy version

You don’t need a new “AI strategy” or a prompt-engineering consultant. You need the boring fundamentals done properly: a site that loads and can be read, honest answers to real questions, consistent facts and people who’ll vouch for you. Do that once, keep it current, and it works for human visitors and AI agents alike.

Frequently asked questions

How do I get my business recommended by ChatGPT?

Make sure ChatGPT’s crawlers can access your site, publish clear factual information about what you offer, answer common buyer questions directly on your pages, keep your details consistent across the web, and earn genuine reviews and third-party mentions. There’s no paid way to guarantee a recommendation.

What is generative engine optimisation (GEO)?

GEO, sometimes called AI search optimisation or answer engine optimisation, is the practice of making your content easy for AI systems to find, understand and cite in their answers. It overlaps heavily with good SEO: crawlable pages, clear structure, direct answers and genuine authority.

Do AI assistants show ads?

Some are starting to experiment with sponsored placements. But the recommendation itself is meant to be earned on relevance, because that’s what makes people trust the assistant. OpenAI, for example, has said the product results ChatGPT shows for shopping are not sponsored.

Should I block AI crawlers from my website?

It depends on your business. Publishers who sell their content may want to block or charge AI crawlers. Most businesses that sell products or services want to be found and recommended, so blocking AI crawlers usually works against them. Check your CDN, because some now block these crawlers by default.

Curious how your business shows up when someone asks an AI for a recommendation? Ask us to take a look. We’ll run the questions your buyers ask and tell you what we find.