There’s a reasonable-sounding idea going around that you can “optimize for AI” the way you once optimized for Google. It would be convenient if that were true. It isn’t. ChatGPT, Google’s AI answers, Perplexity, Claude, and Grok don’t share a brain or a backend. They pull from different indexes, favor different kinds of sources, and reach different audiences, so work that gets you cited in one can do almost nothing for another.
It’s also a moving target. The sources these tools lean on have shifted hard over the past year, sometimes in a matter of weeks. So the real skill isn’t finding the one trick that works everywhere. It’s understanding how each engine actually finds information, and then deciding where your attention is worth spending.
Which engine reads which index
The first thing to get straight is that these tools are not all searching the same web.
ChatGPT and Microsoft Copilot both run on Bing’s index. If your site isn’t in Bing, you’re largely invisible to both, at least for now. That side has enough to it that we cover it on its own, in our piece on showing up in ChatGPT.
One nuance that trips people up: Copilot even lets you answer with Claude as the model now, not just OpenAI’s, but it still pulls from Bing either way. The model doing the talking is separate from the index it draws on.
Google’s AI products, the AI Overviews at the top of results, the newer AI Mode, and the Gemini app, all draw from Google’s own index. That sounds familiar, except the rules for getting cited there are no longer the same as the rules for ranking.
Perplexity runs its own crawler and index. It references Google and Bing as inputs, but it doesn’t depend on either, so being in Bing won’t get you into Perplexity. What matters is whether Perplexity’s crawler has reached your pages.
Claude runs on Brave Search, which maintains its own independent index rather than borrowing Google’s or Bing’s. Anthropic confirmed Brave as the backend, and one analysis found Claude’s citations match Brave’s top results about 86.7% of the time. Brave’s index also feeds Meta AI and Perplexity, so it quietly powers more than just Claude.
Grok is the odd one out. It pulls from the live web and X at the same time, which tilts it heavily toward whatever is fresh and whatever is being discussed on X right now.
Five tools, at least four different sources of truth. That is the whole reason “optimize for AI” is the wrong frame.
The same business shows up at very different rates depending on the engine, and which engine favors it varies from one business to the next. A category that Google’s AI answers cover heavily can be nearly silent on ChatGPT, and the reverse is just as common.
What actually moves you up in each one
Knowing the index is half of it. The other half is what each engine rewards, and it varies.
For ChatGPT and Copilot, the entry requirement is being indexed in Bing, after which clear, well-structured pages do the work. For Google’s AI surfaces, ranking and citation have pulled apart. Ahrefs found that the share of AI Overview citations from organic-top-10 pages fell from 76% in 2025 to 38% by early 2026, so being on page one is no longer enough, or even required. What gets pulled now is content written as self-contained answers, with clear signals about who you are and visible expertise. Perplexity favors fresh, source-rich pages and leans on community content. Claude rewards whatever ranks in Brave, which tends to surface independent and long-tail sources that Google and Bing bury. Grok rewards recency and an active, credible X presence more than an old, authoritative domain.
The levers only rhyme. They aren’t the same, and the same effort pays off differently depending on the tool.
Where the information comes from has shifted, and keeps shifting
This is the part that catches people who set an AI strategy six months ago and assumed it would hold.
The sources these tools cite have moved dramatically. In late 2025, Reddit’s ChatGPT citation share collapsed from roughly 60% to about 10% of responses in about two weeks. Over a similar stretch, YouTube overtook Reddit as the single most-cited source, appearing in around 16% of AI answers to Reddit’s 10%. And the same source can behave completely differently from one engine to the next. Reddit accounts for roughly 44% of the social citations in Google’s AI Overviews but only about 5% in Gemini, two products from the same company.
A few patterns are worth holding onto even as the exact numbers move. Wikipedia carries heavy weight in ChatGPT. Reddit is large in Perplexity and Google’s AI Overviews but unreliable in ChatGPT and nearly absent in Gemini. YouTube is rising across the board. LinkedIn articles have climbed quickly and matter most for B2B. And brands listed across review platforms like G2, Capterra, and Trustpilot see a threefold lift in citations.
The lesson isn’t to chase any one platform. It’s that a presence built only on your own website leaves most of the citation surface untouched, and that whatever mix works today will look different in a few months. A point-in-time audit ages badly, which is why this is worth watching week to week rather than auditing once a quarter.
Slow brains, fast eyes
It helps to separate two things that get blurred together. A model’s base knowledge, the part baked in during training, updates slowly and on a delay, which is why an assistant can confidently describe a version of your business that is a year out of date. The web-search layer bolted on top of most of these tools is a different system, and it moves fast. When one of them runs a live search, it can find and cite content published days ago, as long as that content is indexed and reachable by the right crawler.
You can see the gap in how ChatGPT treats Reddit. Reddit shaped much of what ChatGPT “knows” because it was heavily represented in training, yet it has become an unreliable live citation source for ChatGPT after changes to how it can reach that content. The training memory and the live citation are two separate doors. Fresh, well-structured, crawlable content is what gets you through the fast door, and that door is the one moving the needle right now.
What helps no matter which tool
Underneath the differences, a handful of things help almost everywhere, because they match how all of these systems work rather than any one index.
Being reachable comes first, since each engine has its own crawler and a blocked one makes everything else moot. From there, a few things recur. Content written as clear, self-contained answers that can be lifted out without their surrounding context. Strong signals about who you are, so the machine can identify you confidently, through consistent details, structured data, and recognized profiles like Wikipedia and Wikidata. Third-party mentions and citations, because every one of these tools weighs what other credible sources say about you, not only what you say about yourself. And freshness, which is a thumb on the scale everywhere and decisive in some.
None of that is a trick. It’s the cumulative work that makes you legible to machines that read very differently from people.
How this ties back to traditional search
It would be easy to read all this as “SEO is dead.” It isn’t, and that framing gets businesses in trouble. Several of these engines still lean on traditional search underneath. ChatGPT and Copilot ride on Bing’s index. Google’s AI surfaces, even as citation pulls away from raw ranking, still draw from the same Google index your organic work feeds. Good traditional SEO, clean technical health, earned authority, and useful content is the ground the AI layer stands on. If anything, the shift has made the case for solid SEO stronger, not weaker.
What’s changed is that it is no longer sufficient on its own. Being indexed in Google does nothing for Claude. Ranking first does less for AI Overviews than it once did. Traditional SEO is now the floor rather than the ceiling. You still need it, and you need a layer on top that accounts for how each engine sources and cites.
So where should you actually spend?
There are two reasonable strategies, and the right one depends on your business.
The first is to cover everything: get indexed across Bing, Brave, and Google, keep a credible presence where citations are pooling, and monitor the major engines. That fits when your audience is broad or when you can’t yet say where your buyers do their research.
The second is to optimize for where your audience already is, because the reach is lopsided and shifting. By web-visit share, ChatGPT still leads at roughly half of generative-AI traffic, Google’s Gemini has surged to about a quarter, and together they hold the large majority, with Perplexity, Claude, Grok, and the standalone Copilot app splitting a smaller, more specialized remainder. For most businesses that means ChatGPT and Google’s AI surfaces come first, simply because that’s where the volume is. From there it depends on fit: Copilot and Claude skew toward business and enterprise users, Grok toward an X-native audience, and Perplexity toward research-heavy, high-intent users. A B2B software company and a local services business should not be making the same bets.
Why any of this matters comes down to one number. In Google’s AI Mode, by one analysis, roughly 93% of searches end without a click to any website. When the answer is the destination, the shortlist gets built inside it. A buyer who never clicks through still walks away with three names, and yours is either one of them or it isn’t. That is a pipeline question long before it is a traffic question.
The one thing that won’t change
Most of the specifics here will be dated by the time you’ve settled on a plan. The mix of sources will shift again, a platform will gain or lose ground, and a crawler rule will quietly change what’s visible. What won’t change is the shape of the problem: several engines, reading several different webs, all moving at once. That makes this ongoing work rather than a project with an end date, not a switch you flip once.
Where to start
If you can’t yet say which of these tools surface your business and which describe it in someone else’s words, that’s the place to begin. We track this daily across the accounts we manage, and we can point the same view at yours: where you show up across ChatGPT, Google’s AI answers, Perplexity, and the rest, where you’re missing, and where you’re being represented secondhand. From there it’s a decision about whether to cover everything or concentrate on where your buyers actually are, and we can help you make it.
If you want to see where you stand across all of them before you decide, that is the starting point for the AI visibility work we do.
Sources
OpenAI’s ChatGPT Search relies on Bing’s index (The Keyword)
How Perplexity’s own crawler and index work (AI+Automation)
Claude web search runs on Brave Search, 86.7% citation overlap (Profound)
Brave’s index also powers Meta AI and Perplexity (Rapid Indexer)
How Grok sources from the live web and X (Ansly)
Google AI Overview source selection and the Ahrefs top-10 decline (ZipTie)
The State of AI Citations 2026, Reddit and review-site data (5W via Yahoo Finance)
YouTube overtakes Reddit as the top AI citation source, Tinuiti Q1 2026 data (PikaSEO)
AI chatbot market share, June 2026 (Momentic)
AI search statistics, including AI Mode zero-click rate (Exposure Ninja)
