Platform-Specific Playbooks
Not all AI engines are created equal. ChatGPT, Perplexity, Google AI Overviews, and Microsoft Copilot each have different retrieval mechanisms, content preferences, and citation patterns. Optimizing for one does not guarantee visibility on the others: Profound found only 11% of cited domains overlapped between ChatGPT and Perplexity when it ran the same 100,000 prompts through both.
This chapter gives you a separate playbook for each of the four platforms, then a unified strategy for citation coverage across all of them. The same principles apply to Claude, Gemini, and Google AI Mode, which I note where they differ.
No platform publishes its citation algorithm. The retrieval facts below come from each operator's documentation; the "preferences" tables are my working observations from B2B tracking, not published weights.
The Platform Landscape
Before the individual playbooks, here is what makes each platform different at a fundamental level.
AI Search Platform Architecture
==================================
ChatGPT Perplexity
+------------------+ +------------------+
| Training Data | | Real-Time Web |
| + ChatGPT Search | | Search + Index |
| (OAI-SearchBot) | | (PerplexityBot) |
| | | + LLM Synthesis |
+------------------+ +------------------+
| Tends to favor: | | Tends to favor: |
| Depth, Authority,| | Recent, Sourced, |
| Clear Structure | | Clear Passages |
+------------------+ +------------------+
Google AI Overviews Microsoft Copilot
+------------------+ +------------------+
| Google Search | | Bing Index |
| Index (Googlebot)| | + Microsoft 365 |
| + Gemini model | | Integration |
+------------------+ +------------------+
| Tends to favor: | | Tends to favor: |
| Pages that rank | | Bing-Ranked, |
| well in Google | | Enterprise Docs |
+------------------+ +------------------+
Playbook 1: ChatGPT Optimization
ChatGPT is the largest AI assistant by user base: 900 million people use it every week, OpenAI reported in February 2026. It is also the biggest source of AI referral traffic, at 87.4% of the total in Conductor's 2026 benchmarks report. Optimizing for ChatGPT starts with how it retrieves information.
How ChatGPT Selects Sources
ChatGPT uses two main information pathways:
- Training data: The model learned from a large snapshot of the web. If your content was authoritative and widely referenced before the training cutoff, the model may know your brand without searching. These answers usually carry no links.
- ChatGPT search: When the model decides it needs current information, or the user asks it to search, it retrieves web pages and cites them. Pages reach this index through OpenAI's search crawler,
OAI-SearchBot. OpenAI states that sites that block OAI-SearchBot are not shown in ChatGPT search answers.
Two other OpenAI agents are easy to confuse with it. GPTBot collects training data, and ChatGPT-User fetches a page when a user's request needs it. Neither controls whether you appear in ChatGPT search.
ChatGPT Content Preferences
| Factor | What I observe | Priority |
|---|---|---|
| Content depth | Comprehensive, in-depth content over surface-level pieces | Critical |
| Data specificity | Content with specific, sourced numbers and metrics | High |
| Author authority | Content from recognized domain experts | High |
| Structured formatting | Clear headings, tables, and lists | High |
| Recency | Newer content, especially for technology topics | Medium-High |
| Brand reputation | Brands with many authoritative references across the web | Medium |
| Balanced tone | Neutral, informational content over promotional material | High |
ChatGPT Optimization Tactics
Allow OAI-SearchBot. Check robots.txt and your CDN's bot-protection settings. This is the one technical step with documented, direct impact.
Write definitive guides. In my experience, ChatGPT cites comprehensive resources that cover a topic end to end more often than scattered short posts on the same subtopics.
Lead with the answer. Do not bury your key insights behind lengthy introductions. Put the most citation-worthy information in the first two paragraphs of each section.
Include specific, sourced data points. Instead of "Acme handles a large number of identities," write "Acme manages authentication for [N] million identities in [N] countries," with numbers you can prove. Research supports this: adding statistics raised visibility in generative answers in the GEO paper from Princeton University and collaborators.
Build entity recognition. The more consistently your brand appears across authoritative content on the web, the better ChatGPT knows it, including in answers given without search. Guest posts, industry reports, and analyst mentions all contribute.
Test your ChatGPT visibility regularly. Ask the questions your buyers would ask, in a fresh session, and note whether your brand appears. Ask variations of each question, because answers vary from run to run. Track changes over time as you implement optimizations.
Playbook 2: Perplexity Optimization
Perplexity operates differently from ChatGPT. It searches the web for nearly every query, then synthesizes an answer with numbered inline citations.
How Perplexity Selects Sources
Perplexity retrieves sources in real time for each query. In general terms, it:
- Runs one or more web searches based on the user's query
- Retrieves and ranks candidate pages
- Extracts relevant passages from each source
- Synthesizes an answer with numbered inline citations
- Displays source links prominently
Its index is built by PerplexityBot, which Perplexity says is used to surface and link websites in search results, not to train foundation models.
Perplexity Content Preferences
| Factor | What I observe | Priority |
|---|---|---|
| Content recency | Recently published or updated content | Critical |
| Source attribution | Content that cites its own sources | High |
| Clear structure | Headings, tables, and organized sections | High |
| Topical relevance | Tight focus on the query topic, not broad overviews | High |
| Domain authority | Established domains with strong link profiles | Medium-High |
| Load speed | Fast, server-rendered pages | Medium |
| Original reporting | First-party data and research over aggregation | High |
Community sources matter on Perplexity too. Reddit was its single most cited source, at 6.6% of its citations, in Profound's analysis of 680 million citations.
Perplexity Optimization Tactics
Update often, and update for real. In my experience, recency carries more weight on Perplexity than anywhere else. A current post often beats an older, more comprehensive guide. Keep a consistent cadence of genuine updates.
Optimize for passage extraction. Perplexity extracts short, self-contained passages. Write paragraphs that contain complete, quotable statements without needing context from the paragraphs around them.
Cite your own sources inline. When you use industry data, link to the original source. Citing sources was one of the edits that raised visibility in the GEO paper, whose experiments included a test on Perplexity itself.
Use descriptive subheadings. If a buyer asks "How does CIAM handle passwordless authentication?", a subheading of "How CIAM Handles Passwordless Authentication" makes the matching passage easy to find.
Make pages fast and readable without JavaScript. Real-time retrieval rewards pages that return their content quickly in plain HTML. Meet Google's Core Web Vitals thresholds as a practical baseline.
Playbook 3: Google AI Overviews Optimization
Google AI Overviews run on Google's existing search infrastructure, so your traditional SEO performance directly influences your AI Overview visibility. Google's AI Mode, a separate conversational tab, works from the same index and follows the same rules.
How Google AI Overviews Select Sources
Google's AI features are generated by a Gemini model working from Google's search index. Google states there are no additional requirements to appear in AI Overviews or AI Mode. A page must be indexed and eligible to show with a snippet, and no special schema is needed. In practice, the process looks like this:
- Google retrieves relevant indexed pages for the query (and related sub-queries)
- A Gemini model synthesizes an answer from selected pages
- Sources are linked from the overview
- Clicks and impressions are counted in Search Console as normal "Web" search traffic
Google AI Overview Content Preferences
| Factor | What I observe | Priority |
|---|---|---|
| Traditional search ranking | Pages that already rank for the query or its sub-questions | Critical |
| Snippet eligibility | Content that can be shown as a snippet (no nosnippet) |
Critical |
| Content structure | Clear question-and-answer formatting | High |
| Page authority | Strong domain and page-level links | High |
| Content freshness | Genuinely updated content for fast-moving topics | Medium-High |
| Mobile rendering | Clean mobile pages (Google indexes mobile first) | Medium |
| Schema.org markup | Helpful for clarity, but not required | Medium |
Google AI Overview Optimization Tactics
Rank first. The most reliable path into AI Overviews is ranking well for the query and its likely sub-questions. Classic SEO is the entry ticket.
Use question-based H2 headings. Format subheadings as the questions people search, such as "What is the difference between CIAM and IAM?", followed by a direct, concise answer paragraph.
Keep schema honest. FAQPage and Article markup still help Google understand a page. But since 2023 Google shows FAQ rich results mainly for authoritative government and health sites and has deprecated HowTo rich results. Do not expect markup alone to win a citation.
Create comparison content. AI Overviews often appear for comparison queries ("X vs Y", "best tools for Z"). Build well-structured comparison pages with tables that answer these queries directly.
Google's AI features are the most closely tied to traditional SEO performance. If you do not rank for a query or its sub-questions, you are unlikely to be cited for it. This is where the SEO-to-GEO bridge from Chapter 2 becomes critical.
Playbook 4: Microsoft Copilot Optimization
Microsoft Copilot is built into Windows, Edge, Bing, and Microsoft 365. For many enterprise users it is the AI assistant they meet most often during daily work.
How Copilot Selects Sources
Copilot grounds its web answers in Bing's index. In Microsoft 365, it can also draw on an organization's own data, such as files and email the user can access, which makes it unusual among AI platforms. For your public content, Bing indexing is the entry ticket.
Copilot Content Preferences
| Factor | What I observe | Priority |
|---|---|---|
| Bing search ranking | Pages that rank well on Bing | Critical |
| Enterprise relevance | Content targeting business and enterprise audiences | High |
| Microsoft ecosystem alignment | Content referencing Microsoft technologies, where relevant | Medium |
| Structured data | Schema.org and clear metadata | Medium |
| Professional tone | Business-oriented language and formatting | Medium |
| Bing Webmaster Tools | Verified sites with submitted sitemaps | High |
Copilot Optimization Tactics
Optimize for Bing specifically. Many B2B marketers ignore Bing because of its smaller consumer share. Copilot's reliance on Bing's index makes Bing SEO part of enterprise AI visibility.
Use Bing Webmaster Tools and IndexNow. Verify your site in Bing Webmaster Tools and submit your sitemap. Use IndexNow to notify Bing when pages change, so updates reach the index quickly.
Create enterprise-focused content. Content that addresses enterprise use cases, compliance requirements, and large deployments matches what Copilot users tend to ask.
Reference Microsoft technologies where relevant. If you integrate with Azure, Microsoft Entra ID (formerly Azure Active Directory), or Microsoft 365, say so clearly. Microsoft-related questions are common in Copilot, and clear integration pages answer them.
Unified Cross-Platform Strategy
Optimizing for four platforms separately is impractical for most teams. Here is a unified view of the common ground, based on my experience:
| Optimization Action | ChatGPT | Perplexity | Google AI | Copilot |
|---|---|---|---|---|
| Allow the platform's search crawler | CRITICAL (OAI-SearchBot) | CRITICAL (PerplexityBot) | CRITICAL (Googlebot) | CRITICAL (Bingbot) |
| Comprehensive depth | YES | Partial | YES | YES |
| Recent, genuine updates | YES | CRITICAL | YES | YES |
| Schema.org structured data | Moderate | Moderate | YES | Moderate |
| Author credentials and bios | YES | YES | YES | Moderate |
| Tables and structured comparisons | YES | YES | YES | YES |
| Original data and metrics | CRITICAL | YES | YES | YES |
| Question-answer formatting | YES | YES | CRITICAL | YES |
| Fast page load speed | Moderate | CRITICAL | YES | Moderate |
| Bing Webmaster Tools and IndexNow | Helpful | N/A | N/A | CRITICAL |
| Mobile optimization | Moderate | Moderate | CRITICAL | Moderate |
The 80/20 Rule for Cross-Platform Optimization
In my experience, these five actions cover most of the opportunity across all four platforms:
- Let each platform's search crawler in, and confirm your CDN does not block it
- Publish comprehensive, data-rich content with specific metrics and original research
- Structure content with clear headings, tables, and self-contained sections that AI engines can extract cleanly
- Update content regularly, with visible timestamps that match real changes
- Build and display authority signals through author bios, credentials, and company track record
Do not try to game any platform's citation algorithm. AI engines change their retrieval methods often and rarely announce it. The safest long-term strategy is to be the most authoritative, complete, and current source on your topic. In my experience, tactics that exploit a specific quirk stop working quickly.
Monitoring Platform-Specific Performance
Set up a monitoring cadence to track your visibility across each platform:
Weekly Monitoring Checklist
============================
[ ] Ask 5 key buyer questions on ChatGPT
(note: cited / not cited / competitor cited)
[ ] Ask the same 5 questions on Perplexity
(note: your source cited / not cited)
[ ] Search the same queries on Google
(note: AI Overview shown / your page linked)
[ ] Ask the same questions in Copilot
(note: cited / not cited; check Bing ranking)
Monthly Analysis
============================
[ ] Compare citation frequency across platforms
[ ] Review Search Console's Generative AI report
[ ] Identify platform-specific gaps
[ ] Prioritize optimization based on buyer usage
[ ] Update content based on findings
For a market analysis of GEO across platforms, see GEO Market Research and Industry Analysis. For each engine's documented grounding and citation style, see the engine profiles on GEO Compass.
The next chapter explains the underlying technology, Retrieval-Augmented Generation, so you understand how AI engines find, process, and cite your content.