Which AI platforms should an AEO strategy target?

Three variables decide the target list: how many of your buyers actually use a platform, whether that platform cites external websites in its answers, and whether you can influence what it retrieves. Scored that way, the field sorts into a clear hierarchy. Google's answer surfaces and ChatGPT sit at the top because they combine enormous reach with documented, controllable citation mechanics. AI Overviews reach more than 2 billion monthly users [1], AI Mode passed 1 billion monthly users within roughly a year of launch [2], and ChatGPT hit 900 million weekly active users in February 2026 [3]. No other platform comes within an order of magnitude of that combined footprint.

The citation question filters the list further, because raw reach means nothing for AEO if a platform never links out. Meta AI reported 1 billion monthly users in May 2025 [15], yet its assistant lives inside Facebook, Instagram, WhatsApp, and Messenger, where conversations rarely surface cited web pages. Perplexity is the opposite case: a far smaller user base [10], but a product built around linking to the pages behind its answers [9], which makes each Perplexity user disproportionately valuable as a citation and referral source.

Retrievability is the variable you control directly. Google states that appearing in AI Overviews and AI Mode requires nothing beyond standard Search indexing and snippet eligibility [5]. OpenAI, Perplexity, and Anthropic each publish named crawlers with separate robots.txt controls for search visibility versus model training [8][9][13], so a single overzealous firewall rule or CDN bot filter can silently remove a site from several answer engines at once.

One caution frames everything below: ChatGPT's share of generative AI site visits fell from about 76% to roughly 53% in the twelve months to May 2026 while Gemini and Claude absorbed most of the difference [4]. Every figure in this answer carries its reporting date for that reason, and any prioritization built on these numbers should be revisited quarterly.

Google runs three overlapping answer surfaces, and together they form the largest AEO target by a wide margin. The efficient part is that one body of work covers all three: Google states there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary" [5]. If a page is indexed and eligible to show a snippet in regular Search, it is eligible for Google's AI answers [5].

AI Overviews

AI Overviews are the summaries placed on top of standard results, and they reached more than 2 billion monthly users across 200-plus countries as of July 2025, the most recent figure Google has disclosed [1]. Because the feature sits inside default Google Search, it touches more of the general population than every other answer engine combined. The work here is ordinary technical SEO: keep pages indexed and snippet-eligible, then structure content so a short passage can answer the question outright. Google also notes its AI features display "a wider and more diverse set of helpful links" than classic results, which opens citation room for pages that never won a top-three blue link [5].

AI Mode

AI Mode is Google's conversational search tab, and it passed 1 billion monthly users in May 2026 with query volume more than doubling every quarter since launch [2]. It works through query fan-out, issuing multiple related searches across subtopics to assemble one answer [5]. That mechanic rewards sites that cover a topic in depth rather than a single page chasing a single keyword, because a fanned-out query set can pull several of your pages into the same response.

The Gemini app

The Gemini app passed 750 million monthly active users in February 2026 [6]. It is a chat assistant rather than a search product, and Google notes that only some Gemini responses are grounded on Search results [7], so citation opportunities are less predictable than in AI Overviews or AI Mode. Gemini visibility rides on the same Google Search foundation, which means it requires no separate program, but you should expect less consistent attribution from it.

ChatGPT reached 900 million weekly active users in February 2026, up from 800 million the previous October [3]. Its share of generative AI website visits has fallen, from about 76% in June 2025 to roughly 53% in May 2026, but that decline reflects a growing market rather than a shrinking product: absolute visit volume held flat at the largest count of any assistant [4].

Two mechanics make ChatGPT a controllable target rather than a black box. The first is a dedicated search crawler. OpenAI documents that "OAI-SearchBot is used to surface websites in search results in ChatGPT's search features" [8]. Allowing it in robots.txt governs whether your pages can appear in ChatGPT's cited answers, and it is a separate control from GPTBot, which collects content for model training [8]. A site can decline training and remain fully visible in ChatGPT search. The second mechanic is live retrieval: an agent called ChatGPT-User fetches pages in real time when a user asks about them [8], so fast, accessible pages get represented accurately at the moment of the question.

Citations inside ChatGPT are still the exception rather than the rule, but the trend is steep. Citation presence in US prompts rose from about 1.6% in June 2025 to roughly 6.8% in May 2026, and citation rates run above 20% in categories such as travel and automotive [4]. After a May 2026 interface update, ChatGPT also began sending a much larger share of its referrals to brand homepages, roughly 62 to 63% of referral traffic, up from the mid-20s [4]. For planning purposes that means verifying OAI-SearchBot is not blocked at the CDN level, and treating your homepage as a landing page for AI-referred visitors who arrive with high intent and no context.

None of these four matches the reach of Google or ChatGPT, so each earns a place on the target list only when its audience matches yours.

Perplexity: small reach, dense citations

Perplexity counts more than 45 million monthly active users and processed 780 million queries in May 2025, the last query figure its CEO disclosed publicly [10]. What it lacks in scale it repays in citation behavior: its crawler documentation states that "PerplexityBot is designed to surface and link websites in search results on Perplexity" and is not used for model training [9]. Users who choose a citation-forward engine tend to be doing verification-minded research, which makes Perplexity a priority for finance, health, technical, and policy topics. Allow PerplexityBot, then confirm your key pages render without heavy JavaScript dependence, since retrieval-based engines quote what they can parse.

Microsoft Copilot: the enterprise default

Microsoft last reported 150 million monthly active users across its Copilot products in October 2025 and did not refresh that figure at its January 2026 earnings report, though paid Microsoft 365 Copilot seats grew 160% year over year to 15 million against an installed base of 450 million paid Microsoft 365 seats [12]. Reach is concentrated where Microsoft already is: corporate Windows, Edge, and Office environments. Copilot Search is grounded in Bing and "cites its sources prominently," with inline links and a viewable list of every source used in an answer [11]. The action is to register with Bing Webmaster Tools and confirm Bing indexes your key pages, because Bing eligibility is the gate to Copilot citations [11]. If you sell to IT, operations, or any Microsoft-centric buyer, Copilot belongs in your top three.

Claude: the fastest riser

Claude grew its share of generative AI site visits from about 2% to nearly 9% in the year to May 2026, the largest proportional gain of any platform Similarweb tracked [4]. Third-party estimates put it around 245 million monthly active users by mid-2026, with usage weighted toward professional work: roughly 45% of Claude's API calls originate from enterprise platforms [14]. Anthropic documents three separate crawlers, and the search-relevant ones matter here. Blocking Claude-SearchBot "may reduce your site's visibility and accuracy" in search results, and blocking Claude-User prevents retrieval when a user asks about your pages directly [13]. Companies selling developer tools or professional software should treat Claude as a rising second-tier target and verify neither crawler is caught in a generic bot-blocking rule.

Meta AI: reach without a citation surface

Meta AI reported 1 billion monthly active users in May 2025 across Meta's family of apps [15]. That reach is real, but the assistant operates inside social and messaging contexts where answers rarely function as cited, linkable references to external websites. Until Meta ships an answer surface that credits sources the way search-first platforms do, the reasonable posture is monitoring rather than investment.

Citations do not guarantee clicks

Pew Research found that when an AI summary appeared on a Google results page, users clicked a traditional result in only 8% of visits versus 15% without one, and clicked a link inside the summary itself in just 1% of visits [16]. Being cited builds presence in the answer layer, but treating AEO as a traffic channel on par with classic organic search will disappoint. Measure brand mentions and citation share alongside referral visits.

The numbers age in months, not years

ChatGPT added 100 million weekly users in four months [3], and AI Mode went from launch to a billion monthly users in about a year [2]. Any platform ranking, including this one, is a snapshot, so re-verify reach figures each quarter before reallocating budget on top of them.

Attribution undercounts AI influence

Assistants often mention brands without linking, and users who see a recommendation frequently navigate to the site on their own. After ChatGPT's May 2026 interface update, roughly 62 to 63% of its referral traffic landed on homepages [4], and much of the demand it creates arrives with no referrer at all. Standard last-click analytics will file that under direct traffic, so AI-driven demand routinely looks smaller than it is.

Blocking can happen by accident

OAI-SearchBot, PerplexityBot, Claude-SearchBot, and their sibling agents are distinct user agents with distinct purposes [8][9][13]. An aggressive CDN bot rule, or a blanket robots.txt entry written to stop training crawlers, can also remove a site from search surfaces without anyone noticing. Audit robots.txt and firewall rules per agent, and keep the training decision separate from the visibility decision.

It is not a separate program per platform

Google's documentation says standard SEO fundamentals govern eligibility for its AI features [5], and the retrieval-based engines reward the same underlying quality: crawlable pages that answer questions directly and stay current. Targeting a platform mostly means verifying access and measuring presence there, not producing platform-specific content.

It is not a decision about model training

Whether GPTBot or ClaudeBot may use your content for training is a commercial and legal question with its own controls [8][13]. Search visibility runs through different crawlers, and conflating the two can remove a site from answer engines unintentionally.

It is not permanent

The share shifts of the past year [4] would have upended any fixed platform list written in mid-2025. Prioritization is a quarterly review, not a set-and-forget strategy document.

It is not limited to the platforms in this answer

Newer assistants keep arriving, and some will matter. Before adding one to the target list, apply the same test used throughout this answer by checking for verifiable reach among your buyers, a citation surface that credits external websites, and a documented way to influence what it retrieves. A platform that fails any part of that test cannot yet repay a content investment.