People used to find a business through a search box and a page of links. More and more, they ask an assistant instead: which bakery delivers on Sundays, which running shoes suit flat feet, which plumber can come tomorrow. The assistant reads a handful of sites, compares them and answers. Some can already go further and book or buy.
It’s early, and it’s moving fast. In July 2026, 42% of US consumers told Adobe they had used an AI assistant to shop online, and shoppers an AI sent converted 60% better than other traffic (Adobe). When Google shows an AI summary, people click a regular result in 8% of searches, against 15% without one (Pew Research Center).
Being agent-ready means an assistant can do its job with your site. It can find your pages, get in, read them, understand what you offer, trust it and act on it. The good news: this isn’t a second discipline next to SEO. Search engines, answer engines and agents read the same pages. Most of what makes a site agent-ready is the search basics done properly, and the rest builds on top.
Here are the six steps in the order every engine and agent meets your site, with one thing you can check yourself at each.
1. Can they find your pages?
Everything starts with a link. Google and Bing also read sitemaps, but they’re the only ones that say so. OpenAI, Anthropic and Perplexity don’t document using sitemaps at all, so for them a page that nothing links to may as well not exist. Google puts it plainly: “Every page you care about should have a link from at least one other page on your site” (Google).
A sitemap still helps Google and Bing, and it costs little. But Google itself says a well-linked site of “about 500 pages or fewer” may not need one (Google). Links come first.
Check: start on your home page and click through to your three most important pages, such as your best product, your main service and your booking or contact page, using only the normal menus and links. If you can’t get there by clicking, neither can a crawler.
2. Are they allowed in?
Your robots.txt file tells bots where they may go. AI companies now run several bots each, and each does a different job. At OpenAI, GPTBot collects training data, OAI-SearchBot builds ChatGPT’s search results and ChatGPT-User fetches pages when a person asks (OpenAI). Blocking GPTBot keeps your content out of training; it doesn’t take you out of ChatGPT search. Blocking OAI-SearchBot does: OpenAI says such sites “will not be shown in ChatGPT search answers”.
Anthropic splits its bots the same way: ClaudeBot for training, Claude-SearchBot for search and Claude-User for pages a person asks Claude to read (Anthropic). Google’s token for Gemini training, Google-Extended, “does not impact a site’s inclusion in Google Search” (Google).
Two things catch people out. First, a block can come from a place you never edited. CDNs such as Cloudflare offer settings that block AI bots, and Cloudflare can add its own rules to the top of your robots.txt (Cloudflare). Second, robots.txt isn’t a lock. Fetchers that act for a person often don’t read it: Perplexity-User “generally ignores robots.txt rules” (Perplexity), and so do Google’s user-triggered fetchers (Google).
Check: open yoursite.com/robots.txt in a browser. Look for Disallow: / under User-agent: * or under a search bot you want to be found by: Googlebot, bingbot, OAI-SearchBot, Claude-SearchBot or PerplexityBot. If you use a CDN, look in its dashboard for an AI bot setting too.
3. Can they read what they get?
A browser runs your site’s JavaScript before it shows you anything. Many bots don’t. Google renders JavaScript, but when Vercel measured the AI crawlers, “none of the major AI crawlers currently render JavaScript” (Vercel). Anthropic says Claude’s web fetch tool “does not support websites dynamically rendered with JavaScript” (Anthropic). If your prices, stock or opening hours only appear after a script runs, those readers get an empty page. Even Google recommends rendering on the server, because “not all bots can run JavaScript” (Google).
The answer your server gives matters too. A page that says “not found” while answering with a success code is what Google calls a soft 404, and a page that returns a real error is dropped from its index (Google).
Check: open your most important product or service page and view its source (Ctrl+U, or Cmd+Option+U on a Mac). Search the source for the price or a sentence you know is on the page. If it isn’t there, a reader that doesn’t run JavaScript never sees it.
4. Do they understand it?
Once a bot has your page, it has to work out what the page says. Write the facts that matter where people can see them: what you sell, the price, whether it’s in stock, where you are and when you’re open. Bing asks that “Important information is visible on the URL itself” (Bing), and some AI tools read only the visible text and drop hidden markup altogether.
Structured data, the schema.org markup in your page’s code, still has a job. It powers rich results and merchant listings in Google, as long as it matches what’s on the page (Google). We don’t recommend adding it in the hope that AI will quote you more. Google says “there’s no special schema.org markup you need to add” for its AI features (Google).
The same goes for an llms.txt file: we don’t recommend one for visibility, because Google Search ignores it and no other engine says it reads one (Google).
Check: on your top product or service page, read only the visible text. Are the price, availability, hours, phone number and address all there, and do they match your shop feed and your Google Business Profile?
5. Does anything count against you?
Engines also judge whether a page deserves to be chosen. Google uses Core Web Vitals, which measure how fast and stable your pages are for real visitors, in its ranking systems. It also says it will still show “the most relevant content, even if the page experience is sub-par” (Google). Its AI Overviews and AI Mode are “rooted in our core Search ranking and quality systems” (Google), so what works in search works there.
What hurts is anything that looks like manipulation. Google’s spam policies now name “attempting to manipulate generative AI responses” (Google), and Bing warns that prompt-injection content “may result in reduced visibility or removal” (Bing). Be wary of tools that promise hidden text for AI, or a number for where you stand in AI answers. As Google puts it, “No third-party tool has access to our internal ranking or AI systems” (Google).
Check: in Google Search Console, open Manual actions, which should say no issues were detected, and the Core Web Vitals report for mobile and desktop. If you don’t use Search Console yet, setting it up is your first job.
6. Can agents get things done?
This is the newest step. Agents such as ChatGPT’s and Claude’s browsers can now fill in forms and click through a checkout for a person. They stop at the same walls a hurried customer would, and some they won’t climb at all. Claude in Chrome won’t get past CAPTCHAs (Anthropic), and it won’t create accounts or make purchases for you (Anthropic). OpenAI notes that “some sites block automated browsers or require a CAPTCHA”, and that ChatGPT “may not be able to complete a task on those sites” (OpenAI).
So the question is simple. Can someone, or something acting for them, get from your product page to the point of paying or booking without a puzzle or a forced account?
Check: in a private browser window, do your main action as a brand-new customer: buy, book or request a quote. Note every CAPTCHA, every “checking your browser” screen and every step that makes you sign up first.
In short
Work through the steps in order. A page that can’t be found or isn’t let in can’t be read, and a page that can’t be read can’t be understood. Most of the fixes are small: a link in a menu, a line in robots.txt, a price written into the page. Each one helps search engines today and agents tomorrow.
