The moment I realised I was doing AI search wrong
About six months ago, I ran a simple experiment. I asked ChatGPT, Perplexity and Google AI Mode the exact same question: "What are the best SEO strategies for small businesses in 2026?" Then I compared the sources each one cited.
The overlap was almost zero.
ChatGPT pulled from Wikipedia articles and well-structured knowledge bases. Perplexity cited Reddit threads and forum discussions. Google AI Mode leaned heavily on YouTube videos and rich media content. Same question, three completely different source pools. That was the day I stopped treating conversational search as a single channel and started building a platform-specific citation strategy.
The AI search ecosystem has now hit 45 billion monthly sessions, according to Exposure Ninja's AI search report. If you are still treating all AI search engines the same way, you are leaving visibility on the table across every single one of them.
The numbers that changed my approach
Let me lay out the landscape as it stands right now. ChatGPT has 883 million monthly users, commanding 80.49% of the AI search market share. Google AI Overviews reaches 1.5 billion monthly users across its search results. Perplexity, the smallest of the three, still pulls in 22+ million active users who tend to be highly engaged researchers and decision-makers.
Those are massive, distinct audiences. And each platform has wildly different citation preferences. Here is what the data actually shows:
- ChatGPT favours Wikipedia and encyclopaedic content. A staggering 47.9% of its top citations come from Wikipedia-style sources.
- Perplexity heavily cites Reddit, with 46.7% of citations pulling from community discussion threads.
- Google AI Overviews prefers YouTube and multi-modal content, with 23.3% of citations coming from video sources.
These are not small differences. They are fundamentally different content ecosystems. And most businesses I audit are only creating content that works for one of them, if that. This is exactly the kind of gap I explore in our AI SEO hub, where I break down how each platform discovers and ranks content.
How ChatGPT decides what to cite
ChatGPT has a strong preference for authoritative, well-structured, encyclopaedic content. Think of it as a librarian. It wants clean definitions, comprehensive coverage and content that reads like a reference resource.
Here is what I actually do to get cited by ChatGPT:
I write definitive, entity-rich pages. When I create content targeting ChatGPT citations, I structure it almost like a Wikipedia article. Clear headings, factual statements, comprehensive coverage of a topic from multiple angles. I avoid opinion-heavy intros and get straight to the substance.
I build structured data into everything. ChatGPT's training data and retrieval systems respond well to pages with clean schema markup. FAQ schema, HowTo schema, Article schema. All of it helps these models understand what a page is about and whether it is worth citing.
I focus on topical authority. ChatGPT does not just look at a single page. It seems to evaluate whether a site has depth on a topic. This is where semantic SEO becomes critical. One blog post on a topic will not cut it. You need a cluster of interconnected content that proves you know the subject inside and out.
A mistake I made early on was writing content that was too branded and promotional. ChatGPT skips right past that. It wants informational content that answers questions directly. If your site reads like a sales pitch, you will not get cited. I wrote about this pattern in more detail in our ChatGPT ranking insights piece.
What Perplexity actually wants
Perplexity is the complete opposite of ChatGPT in terms of source preference. Where ChatGPT wants polished reference material, Perplexity gravitates toward community-generated content, forum discussions and real-world practitioner insights.
That 46.7% Reddit citation rate tells you everything. Perplexity values authenticity, lived experience and the kind of raw, practical advice that people share in communities.
Here is my Perplexity strategy:
I maintain active Reddit presence in relevant subreddits. Not spamming links. Actually contributing helpful, detailed answers to questions in my industry. Perplexity's crawlers pick up these discussions and frequently cite them as primary sources. The key is being genuinely useful, not promotional.
I write first-person, experience-driven blog content. Perplexity seems to prefer content that includes real examples, case studies and practitioner perspectives. "Here is what happened when I tried X" performs better than "X is a strategy that businesses should consider." This is the kind of content I help clients build through our SEO blog writing service.
I make sure my content answers specific, long-tail questions. Perplexity users tend to ask detailed, specific questions. Content that directly addresses those specific queries, with clear answers up front, gets cited more often. I structure posts with question-based H2s and immediate answers in the first sentence of each section.
If your brand is not showing up in Perplexity results, it is likely because you do not have enough community presence or practitioner-style content. This is a common problem I see when AI is not showing a brand in search results.
Google AI Mode plays by different rules entirely
Google AI Overviews and AI Mode pull from Google's existing index, but they have a clear preference for multi-modal content. That 23.3% YouTube citation rate is significant. Google wants to see that you are creating content across multiple formats, not just text.
What this looks like in practice:
I pair every major blog post with a YouTube video. It does not need to be a high-production video. Screen recordings, talking-head explanations, even slide-based walkthroughs all work. The point is having a video asset that Google can reference alongside the written content. According to DemandSage, multi-modal content continues to dominate AI Overview citations.
I build content that already ranks in traditional Google search. Google AI Mode heavily draws from pages that perform well in organic search. This means all the traditional SEO fundamentals still matter here. Page speed, mobile experience, backlink profile, content depth. You cannot skip the basics and expect AI Mode to find you.
I use structured data aggressively. Google's AI systems are better at parsing structured data than any other platform. VideoObject schema, Article schema, FAQ schema. These all increase the chances of being pulled into an AI Overview. I covered this in detail in our schema markup guide.
I focus on E-E-A-T signals. Google AI Mode cares deeply about expertise, experience, authoritativeness and trustworthiness. Author pages, credentials, cited sources within your content, published bylines on other sites. All of these compound over time. Digital PR is one of the most effective ways to build these signals at scale.
My multi-platform content workflow
Knowing the differences is one thing. Actually building a workflow around them is where most people get stuck. Here is the process I use for every major piece of content I create:
Step 1: Start with a comprehensive, well-structured blog post. This is the foundation. It needs to be thorough enough for ChatGPT's encyclopaedic preferences, with clear definitions and factual depth. At the same time, it should include first-person insights and real examples that Perplexity favours. This dual approach is central to any good AI content strategy.
Step 2: Create a companion video. Even a 5-minute walkthrough of the key points. Upload it to YouTube with proper descriptions, timestamps and tags. This feeds Google AI Mode's multi-modal appetite.
Step 3: Share genuine insights on Reddit. Not a link to the blog post. Instead, write a standalone comment or post that shares a key finding or technique from the content. Let the value stand on its own. If it is good enough, Perplexity will find it.
Step 4: Build supporting content around the topic. Create related FAQ pages, glossary entries and topic cluster pages that reinforce your authority. This helps with ChatGPT's topical authority evaluation and Google's E-E-A-T assessment.
Step 5: Ensure proper technical access. Make sure your robots.txt is configured for AI discovery. If AI crawlers cannot access your content, none of the above matters.
Mistakes I have made along the way
I want to be honest about what did not work, because I think these are common traps.
Trying to game Reddit. I initially tried posting promotional content in subreddits with thin "helpful" wrappers. It got downvoted, removed by moderators and did nothing for Perplexity citations. The only thing that works is being genuinely helpful. People and algorithms can both tell the difference.
Ignoring ChatGPT because "Google is bigger." With 883 million monthly users and growing, ChatGPT is too large to ignore. Especially for B2B queries, where I have seen ChatGPT increasingly become the first place decision-makers go for research. If ChatGPT is not mentioning your brand, you are missing a significant discovery channel.
Creating video content that is just the blog post read aloud. Google AI Mode seems to prefer videos that add something beyond the written content. Demonstrations, screen shares, visual examples. The video needs to provide standalone value, not just be a rehash.
Not monitoring competitor citations. I spent months building content without checking who was actually getting cited in my target queries. When I finally did, I found that competitors were being mentioned instead because they had better community presence and more diverse content formats. That wake-up call reshaped my entire approach.
How to audit your current AI search visibility
Before you build a strategy, you need to know where you stand. Here is my quick audit process:
Test 20 queries across all three platforms. Pick the 20 most important questions your target audience asks. Run them through ChatGPT, Perplexity and Google AI Mode. Note which sources get cited for each query on each platform.
Map your coverage gaps. Where are you cited? Where are competitors cited instead? Which platforms are you completely invisible on? This gives you a clear picture of where to focus first.
Check your content format mix. Do you have text content, video content and community presence? If you are only producing one format, you are structurally unable to get cited across all three platforms.
Review your LLM visibility signals. Are your pages structured in ways that AI models can parse? Do you have clean entity relationships, schema markup and authoritative backlinks? These foundational signals affect visibility across every platform.
The trends covered by Search Engine Journal confirm that enterprise brands are already investing heavily in multi-platform AI search strategies. If you are not doing this yet, you are falling behind.
What I expect to change in the next 12 months
The citation preferences I have outlined are based on current data, and they will shift. Here is what I am watching:
Perplexity is expanding its source base beyond Reddit. I expect its citation profile to diversify, which means building authority on forums alone will not be enough long-term. Your own site needs to be strong enough to get cited directly.
ChatGPT is getting better at finding and citing niche expert content. The Wikipedia dominance will likely decrease as retrieval systems improve. This is good news for smaller, specialist sites.
Google AI Mode will continue favouring its own ecosystem. YouTube, Google Business Profiles and structured data within Google's index will remain the primary citation sources. Playing within Google's ecosystem is not optional here.
The brands that will win are the ones building presence across all three platforms now, before the strategies become obvious and the competition intensifies. Consistent brand mentions across the web will become one of the strongest signals for AI citation across every platform.
The bottom line
You cannot build one type of content and expect to show up across ChatGPT, Perplexity and Google AI Mode. The data makes that clear. Each platform has distinct preferences, distinct source pools and distinct ways of evaluating what deserves to be cited.
The strategy is not complicated, but it does require discipline. Build encyclopaedic reference content for ChatGPT. Create genuine community presence for Perplexity. Produce multi-modal, technically sound content for Google AI Mode. And tie it all together with a solid foundation of topical authority and structured data.
That is what I do. It is working. And if you want help building this for your brand, our AI SEO hub is the best place to start.