Confirm the page is indexed before anything else
Google's AI surfaces run on ordinary Google Search eligibility. Nothing else. Your page has to be indexed and eligible to be shown in Google Search with a snippet. Open Search Console, run URL Inspection on the exact address you want cited, and read the result. If it isn't indexed, stop here. Nothing further down this page matters until that's fixed, and it's an indexing problem, not an AI problem.
The technical requirements come to three, and they're short. Googlebot isn't blocked. The page works, meaning Google only indexes pages served with an HTTP 200 success status code. The page has indexable content, in a file type Google Search supports, that doesn't violate the spam policies. Meet all three and you're eligible. Google's AI features page still says plainly that indexing and serving still aren't guaranteed.
Copilot answers off the Bing index, so ask the same first question there with a different tool. Sign in to Bing Webmaster Tools and confirm the URL is indexed. ChatGPT, Perplexity and Claude publish no index checker at all, so the crawler checks below are the only diagnostic you get on those three.
Check that nothing is suppressing your snippet
A snippet directive can remove you from AI Overviews and AI Mode while your rankings sit untouched, and it's the easiest thing on this list to inherit from somebody else's template. Google's robots meta tag documentation says nosnippet will also prevent the content from being used as a direct input for AI Overviews and AI Mode. A max-snippet value of 0 works the same as nosnippet. Set it to -1 and Google picks the length itself.
View the page source and search it for nosnippet, max-snippet and noindex. Then check the response headers, because the same rules can arrive as an X-Robots-Tag header, which is where nobody thinks to look. The source can read perfectly clean while the server is still sending the directive.
data-nosnippet is the one that hides in a template. It marks a passage, not a whole page. It works on span, div and section elements only and does nothing on any other tag. Somebody wraps a boilerplate block in it, the template moves, and now it's wrapped around the exact paragraph you want quoted. Search your source for data-nosnippet and read what's inside each one.
There is no AI-only opt-out at Google. The directive that pulls you out of AI Overviews and AI Mode also kills your regular search snippet, so it's one decision, not two. If you want to be cited, you want none of these on the page.
Get the bot names right, because the wrong one does nothing
Three of these engines decide search visibility with a crawler whose name isn't the one people block. The ones that govern citation are OAI-SearchBot for ChatGPT, PerplexityBot for Perplexity and Claude-SearchBot for Claude, with Googlebot and Bingbot covering everything else. Open your own robots.txt in a browser right now and read every Disallow line against that list.
OpenAI's documentation spells out the consequence: sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. GPTBot is the training crawler, described as making the foundation models more useful and safe, and blocking it changes nothing about ChatGPT search. OAI-SearchBot didn't exist until late 2024, so a rule written in 2023 could only have named GPTBot. A site that blocked by name is still open to ChatGPT search and thinks it isn't. A site that disallowed every user agent has been shut out since the day the bot launched.
Perplexity says PerplexityBot is designed to surface and link websites in search results on Perplexity and isn't used to crawl for foundation models. Anthropic draws the same line: Claude-SearchBot navigating the web to improve search result quality on one side, ClaudeBot doing the training crawl on the other. Blocking one leaves the other running, so a single blanket rule is almost never what the person writing it meant.
Google-Extended controls nothing about AI Overviews or AI Mode, and most published advice gets this backwards in both directions. Google's crawler documentation says it manages whether your content may be used for training future generations of Gemini models, and that Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search. Block it and you're still in AI Overviews. Allow it and you're not any more likely to get in.
The user-triggered fetchers ignore all of this. ChatGPT-User, Perplexity-User and Claude-User run when a person asks for a specific page. OpenAI says that because these actions are initiated by a user, robots.txt rules may not apply, and Perplexity says its fetcher generally ignores robots.txt rules. An opt-out stops the index. It doesn't stop somebody pasting your URL into a chat.
Check the edge, more than the robots file
robots.txt isn't where blocks actually happen. Your file can say allow while Cloudflare, Akamai or your host's bot management hands the crawler a 403 or a challenge page, and you won't see it from outside. Google lists it in its own best-practice list for AI features: make sure crawling is allowed in robots.txt, and by any CDN or hosting infrastructure.
Pull your server and CDN logs and search the user agent field for the bot names. Then read what you find two ways. A bot that shows up and gets a 403 or a challenge page is being turned away at the edge. That's a firewall rule, and it has nothing to do with your content. A bot that never appears at all hasn't reached you yet, which is more often a discovery or indexation problem than a block, though a WAF that drops requests before logging can look the same. Either way, your CDN's bot-management rules are the next stop. Each vendor publishes its source addresses so you can confirm a hit was genuine: openai.com/searchbot.json, perplexity.ai/perplexitybot.json and claude.com/crawling/bots.json.
Perplexity asks for both things in its own instruction: allow PerplexityBot in robots.txt and permit requests from its published IP ranges. An allow line alone doesn't finish the job if your firewall filters by address. Then test it yourself. Request your own page with curl, setting the user agent to each bot name in turn. A 403 back is a real finding, because a user-agent rule is turning the bot away. A 200 back proves less than it looks. The request came from your address, not the crawler's, so a firewall that filters by IP will wave you through and still block the bot. Only the logs settle that one.
Copilot: pull NOARCHIVE, then ping IndexNow if you want
Two directives govern whether Copilot can cite you, and one of them removes the link itself. Microsoft's September 2023 announcement is the published statement on it, from back when Copilot was called Bing Chat. NOCACHE means the content may be included in answers, with only the URL, snippet and title displayed. NOARCHIVE means the content will not be included in answers and will not be linked to in them. Carry both and Microsoft treats it as NOCACHE.
NOARCHIVE isn't a cache setting, whatever the name suggests. Plenty of sites carry one that somebody added years ago for a reason that had nothing to do with this. Search your source and your headers for it the way you searched for nosnippet. No tag at all is full eligibility, so on this one, doing nothing is the right configuration.
You can only push freshness here, and that isn't a request to be cited. IndexNow takes a GET to the endpoint with a url parameter and a key parameter, the key running between 8 and 128 characters, plus an optional keyLocation when your key file doesn't sit at the site root. A 200 back means the URL was submitted. It promises nothing about whether or when anything gets indexed.
What you can measure, and what nobody reports
Google folds the AI surfaces into normal reporting with no breakout. Its documentation says sites appearing in AI features are included in the overall search traffic in Search Console, reported on in the Performance report within the Web search type. That's all of it. There's no AI Overviews row and no AI Mode filter, so anyone quoting you an AI Overviews click-through rate out of Search Console is reading a number that doesn't exist there.
The other four report no AI citations to you at all. Bing Webmaster Tools shows Bing search impressions and clicks, but Microsoft publishes no Copilot citation breakout, and OpenAI, Perplexity and Anthropic publish no publisher analytics of any kind. Weren't cited? Nobody tells you. Were cited? Nobody tells you that either.
So you check it by hand, on a schedule. Write down the questions a customer would actually type. Run every one in each engine on the same day each month, and record whether you were named and which page got cited instead of yours. Never change the question list. A list that moves gives you a story, not a measurement.
If you're still not being cited
There's no appeal, and it's worth knowing that before you go hunting for one. None of the five has a form, a support queue or an escalation path for inclusion in an AI answer, and nobody accepts the request. One more limit: all of this assumes pages you control. If the answers you want changed are about you personally and they cite sites you don't own, none of these checks touch that. It's separate work.
Run the four checks in order and stop at the first failure. Is the URL indexed in that engine's underlying index? Is a snippet directive suppressing it? Is the named search bot reaching you at all, confirmed in your logs rather than assumed from your robots file? Last, does that bot specifically get a 200. The third one is where most of these end, and it's the only one you can't see from outside your own server.
If all four pass and you're still not cited, there's nothing left to file. None of the five documents how it picks among the pages that are already eligible, so every published method for ranking in Perplexity or getting picked up by ChatGPT is inference, not documentation. Price anybody selling you a guaranteed AI citation accordingly.
What's left is ordinary work, and Google calls it ordinary work, not an AI requirement: internal links so your content is easily findable, a great page experience, and important content available in textual form, supported by images and video where they apply, with structured data that matches the visible text on the page and current Business Profile and Merchant Center information. None of that is AI-specific and none of it is new.