The Platform Fragmentation Problem

One of the most important and overlooked realities of GEO is that different AI platforms cite different sources. Optimizing for one does not guarantee visibility on another.

The 11% Overlap Finding

When Profound ran 100,000 identical prompts through ChatGPT and Perplexity, only 11.0% of cited domains appeared on both platforms (July 2025). The same analysis put overlap between other engine pairs between 6% and 16.4%. A brand that shows up often in ChatGPT answers can be nearly invisible on Perplexity, and vice versa.

A second study points the same way. Ahrefs ran 15,000 long-tail queries through Google and through four AI assistants. Only about 12% of the URLs the assistants cited also ranked in Google's top 10 for the original prompt. Perplexity was the outlier, closest to Google's results.

This fragmentation exists because each AI platform:

• Uses different retrieval systems and web indexes (Bing, Google, or its own crawler)
• Rewrites the user's prompt into different follow-up searches before retrieving
• Weights authority, freshness, and source types differently
• Draws on different training data for answers it gives without searching
• Serves different user bases with different expectations

Platform Citation Behaviors

The behaviors below are my working observations from B2B tracking, anchored to published data where it exists. For each engine's documented grounding and citation style, see the engine profiles on GEO Compass.

Platform	How it retrieves	What it tends to cite (my observation)
ChatGPT	Training data plus live search via OpenAI's own crawler, OAI-SearchBot	Authoritative domains, technical depth, well-structured content
Perplexity	Live web search on nearly every query	Recent, well-structured pages; community sources such as Reddit
Google AI Overviews and AI Mode	Google's search index	Pages that already rank well in Google
Microsoft Copilot	Bing's index	Pages Bing indexes and trusts, including enterprise documentation
Claude and Gemini	Training data plus web search when enabled	Clear, factual, well-attributed pages

ChatGPT

ChatGPT is the largest source of AI referral traffic by a wide margin. Conductor's benchmark found ChatGPT drove 87.4% of AI referral traffic across 10 industries (May to September 2025). When a CISO asks "What are the leading CIAM platforms for financial services?", ChatGPT may answer from training data or run a web search first, then cite the pages it retrieved.

ChatGPT tends to cite:

• Domains with long publishing histories and deep topic coverage
• Content with specific technical details and accurate claims
• Sources that are referenced across multiple other authoritative documents
• Well-structured content with clear heading hierarchies

In Profound's analysis of 680 million citations (August 2024 to June 2025), Wikipedia was ChatGPT's most cited single source at 7.8% of all its citations.

Microsoft Copilot

Copilot is built into Windows, Edge, and Microsoft 365, so many people meet it during active work rather than in a separate research session. Its web answers draw on Bing, so Bing indexing is the entry ticket.

Copilot tends to cite:

• Pages that Bing indexes and ranks well
• Enterprise-focused technical documentation
• Content from established enterprise vendors
• Microsoft Learn and other Microsoft-ecosystem pages for Microsoft-related topics

Google AI Overviews and AI Mode

Google AI Overviews appear directly in search results. Conductor found them on 25.11% of about 21.9 million US Google searches in September and October 2025. AI Mode is Google's separate conversational search tab, built on the same index.

Google's AI features tend to cite:

• Content that already ranks well in traditional Google search
• Pages with clear, extractable answers near the top
• Content matching Google's E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) guidance
• Pages with structured data that matches the visible content

Google states there are no extra requirements to appear in AI features beyond normal Search eligibility. Standard SEO is the entry ticket here.

Perplexity

Perplexity searches the live web for nearly every query, so freshness and crawlability matter more than on any other platform. Profound found Reddit was Perplexity's top single source, at 6.6% of its citations.

Perplexity tends to cite:

• Recently published or updated content
• Well-structured pages with clear answers in the first paragraphs
• Sources with clean HTML and minimal JavaScript rendering dependencies
• Pages that load quickly and are easily parseable

What This Means for Your Strategy

You cannot build a single GEO strategy and expect it to work everywhere. You need to understand where your buyers actually do their research and prioritize accordingly. The persona mapping below is my starting hypothesis; confirm it with your own customer interviews and "how did you hear about us" data.

• Enterprise B2B buyers (CISOs, VPs): In my experience, they lean on ChatGPT and Microsoft Copilot for early vendor research. Start there.
• Technical evaluators (engineers, architects): Often use Perplexity and Claude for quick technical comparisons. Keep technical content current and easy to parse.
• Finance and compliance teams: Still start many regulatory and compliance queries in Google, where AI Overviews appear. Strong classic SEO covers most of this.

Building a Multi-Platform Strategy

Rather than trying to optimize for every platform simultaneously, follow this prioritization approach:

1. Identify your primary buyer persona: Who makes or influences the purchase decision?
2. Map their AI platform usage: Which AI tools does this persona use for research?
3. Prioritize the top two platforms: Focus your initial GEO efforts on these
4. Expand coverage over time: Once you have citation traction on priority platforms, extend to secondary ones