Guide hub Search and AI visibility: what you actually control

You searched your name in Google, then asked ChatGPT the same question and got a different answer citing somebody else. Ranking in a list of links and being quoted in a generated answer are two separate jobs, decided by different crawlers and diagnosed in different places. Nothing on the AI side takes a submission, so the work is access first, then a page worth quoting. Here's the order.

What all of this buys you

Eligibility, never placement. You can make Google, Bing, Brave, ChatGPT, Perplexity and Claude able to reach, index and quote your pages, and that is the entire outcome on offer. Google says on its own AI features page that indexing and serving aren't guaranteed even when you meet every requirement, and OpenAI's complete published statement on ranking ends with placement is not guaranteed. Nobody publishes how they choose among the pages that already qualify.

What it costs you to do alone

An afternoon to open every door, and then it never quite ends. Bing's setup is about twenty minutes. The ChatGPT fix is two lines in robots.txt and about 24 hours before you judge it. Brave takes one URL at a time, with no account and no ticket number. The recurring cost is the measuring: fifteen questions a buyer would actually type, run in each engine on the same day every month, because not one of them reports a citation to you.

The mistake that makes people quit

Your robots.txt says allow and your CDN says 403. A bot rule at Cloudflare or your host turns the crawler away before robots.txt is ever read, nothing announces it, and no console anywhere shows you. People run the whole checklist, stay uncited for months, and decide the category is a scam. Your own access logs are the only proof, and they settle it in ten minutes.

There is no submission form, for any of them

Read this before you buy anything, because it disqualifies most of what gets sold in this category. Google states there are no additional requirements to appear in AI Overviews or AI Mode, and closes the door on the file rituals by name: you don't need to create new machine readable files, AI text files, or markup to appear in these features. An llms.txt is not the price of admission.

The rest are the same answer in different words. ChatGPT publishes no submission tool and no ranking checklist. Perplexity publishes no way to add, suppress or remove a citation. Brave's only self-serve tool re-fetches one URL at a time, with no account, no bulk upload and no confirmation. Bing is the one exception and it is twenty minutes of free setup, not a program you fund.

So there's nothing to apply for, which means anyone selling you AI search submission or ChatGPT ranking factors invented the product. What each engine does publish is the name of the crawler that decides whether you're eligible to be quoted at all, and that name is the whole documented surface. Every guide beneath this page is those same checks run against a different company's bot.

Ranking and being cited are different jobs

Ranking is a position among links, so you're trying to move up a list. Citation is binary: the assistant either names you as a source or it doesn't. A page can sit fifth on Google and still be the page ChatGPT quotes, and it can sit first and never get quoted once. Working on one does not automatically deliver the other.

They also fail differently, which is why people miss the second one entirely. A ranking problem is visible, because Search Console gives you impressions, position and clicks. A citation problem is invisible by design: OpenAI, Perplexity and Anthropic publish no publisher analytics of any kind, and Google folds its AI surfaces into ordinary Performance reporting with no AI Overviews row and no AI Mode filter.

Most sites have never checked either one. They've never read their own robots.txt against the current bot names, and they've never once run the questions their customers actually ask through an assistant to see whose page gets named instead. Both of those are an afternoon, and doing them in that order is the difference between fixing this and guessing at it.

Confirm you're indexed, then get the bot names right

Google's AI surfaces run on ordinary Google Search eligibility and nothing else, so if the URL isn't indexed, nothing further down this page matters. Run URL Inspection in Search Console on the exact address you want cited. Copilot answers off the Bing index, so ask the same first question in Bing Webmaster Tools. ChatGPT, Perplexity and Claude publish no index checker at all, which is why the crawler check below is the only diagnostic you get on those three.

Then open your own robots.txt and read every Disallow line against the names that actually govern citation: OAI-SearchBot for ChatGPT, PerplexityBot for Perplexity and Claude-SearchBot for Claude, alongside Googlebot and Bingbot. The training crawlers are separate bots, GPTBot and ClaudeBot, and blocking those costs you nothing in search. Google-Extended governs neither: blocking it won't remove you from AI Overviews and allowing it won't get you in.

Brave is its own case and people assume it isn't. Brave runs its own index and serves results solely from it, so your Google position counts for nothing there and a removal you won at Google removed nothing from Brave. Its crawler doesn't advertise a differentiated user agent, which has one useful consequence: the Googlebot rule in your robots.txt is the Brave rule, and there's no Brave line to add.

The block is almost never in the robots file

This is the step that eats months. Your file can say allow while Cloudflare, Akamai or your host's bot management hands the crawler a 403 or a challenge page, and none of that is visible from outside. Google puts it in its own best-practice list: make sure crawling is allowed in robots.txt, and by any CDN or hosting infrastructure. Perplexity goes further and asks you to permit its published IP ranges as well.

Pull your server and CDN logs and search the user agent field for each bot name. A bot that shows up and gets a 403 or a challenge page is being turned away at the edge, which is a firewall rule and has nothing to do with your content. A bot that never appears hasn't reached you. Curling your own page with the bot's user agent proves less than it looks, because the request comes from your address rather than the crawler's, so a firewall filtering by IP waves you through and blocks the bot anyway.

Read the page for snippet directives while you're in there, because they suppress citation without touching your rankings. nosnippet removes your content as a direct input to AI Overviews and AI Mode, max-snippet set to 0 does the same thing, data-nosnippet can be sitting around one paragraph inside a template, and all of it can arrive as an X-Robots-Tag header where nobody thinks to look. For Copilot the equivalent is NOARCHIVE, which despite the name is not a cache setting and removes the link entirely.

Make the page legible, then make it worth quoting

Access makes you eligible. What gets you quoted is a page that answers the question in text a machine can lift. Structured data is how a model resolves who you are: Organization markup on the homepage carrying an @id, and sameAs pointing at your LinkedIn company page, your Wikidata item and anywhere else your name is already published. That's what collapses five loose references into one entity.

Schema has never been a ranking factor and nobody should sell it to you as one. It buys eligibility and legibility, not position. Two types stopped paying in August 2023, when Google cut FAQ rich results back to government and health sites and retired HowTo entirely. And the common failure isn't a missing tag: it's a plugin writing one Organization block while your theme writes a second one with a different name and logo, which Google resolves by trusting neither.

Then the ordinary part, which Google lists as ordinary work rather than an AI requirement: internal links so your content is findable, a good page experience, important content available in textual form, and structured data that matches the visible text. Correct markup over a thin page describes a thin page accurately, and that is all it does. Write the answer first, mark it up second.

Measure it by hand, because nobody reports it

There's no Search Console for AI citation. Google includes its AI surfaces in overall search traffic with no breakout, so anyone quoting you an AI Overviews click-through rate out of Search Console is reading a number that doesn't exist there. Bing gives you search impressions and clicks with no Copilot citation breakout, and OpenAI, Perplexity and Anthropic publish nothing to publishers at all.

You get two real instruments and both are worth wiring up before you change anything. ChatGPT appends utm_source=chatgpt.com to referral URLs, so build a GA4 segment on that source today and treat the number as a floor, since a citation that gets read and not clicked never reaches you. Bing's AI Performance report counts citations from your site shown as sources in AI answers, which is worth opening monthly and not worth tuning for, because no control in that tool moves it.

Everything else is a hand-run test, and it only works if you keep it identical. Write down fifteen questions a buyer 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 list. A question list that moves gives you a story instead of a measurement.

What none of this moves

Eligibility is a precondition, not a result. Pass every check on this page and an engine can still name your competitor, and there's nowhere to appeal it: none of these assistants offers a form, a support queue or an escalation path for being included in an answer, and not one of them documents how it picks among the pages that already qualify. Price anybody selling a guaranteed AI citation accordingly.

All of it also assumes pages you control. If the answer you want changed is about you personally and it's citing sites you don't own, none of these checks reach it. What moves that is the cited page itself getting corrected, delisted or removed at the publisher, and then publishing accurate material for a model to read instead. That's different work, with different rules and a different timeline.

And nothing transfers between engines. A delisting you won at Google removed nothing from Brave, which runs a separate index, takes a separate email, and says a delisting it agrees to can take up to 30 days. We'll tell you which of those situations you're actually in before you pay, because it changes what you should buy.

Access is free. Being the answer isn't.

Everything above you can run yourself this week, and you should, because none of it costs money. What takes somebody's ongoing attention is keeping five crawlers reaching a site that changes every week, reading the logs when one of them stops, and writing pages good enough that an assistant quotes them instead of your competitor. That's what our SEO and AI Optimization service does: we record what Google, Bing, Brave and the assistants say about you today, fix the site and the sources behind them, then ask the same questions again so you can see what moved. It's quoted against your situation rather than sold at a list price.

Have us do it.

Everything above, filed for you, chased for you, and reported back. One flat fee.

See the service

Every guide we publish is free and ungated. Browse all of them, see what we do and what it costs, or read why we built this company.

Written by Drew Chapin, who ran all of this on his own name first.