How to Appear in ChatGPT Recommendations: What Our Research Shows Actually Works

If you search for "how to appear in ChatGPT" right now, most of what you will find is speculation. Publish more content. Add schema markup. Build backlinks. Optimise for E-E-A-T. These are SEO tactics repackaged with a new label, and our research suggests most of them have weak or no correlation with whether AI assistants actually recommend your brand.

We have been studying how ChatGPT, Gemini, Perplexity, Google AI Overviews, and Google AI Mode select and cite brands across thousands of real queries. The findings paint a different picture from what most of the industry is publishing. Here is what actually works – backed by citation data, not assumptions.

The Two Layers of AI Knowledge

Before you can improve your visibility in ChatGPT or any AI assistant, you need to understand how these systems form their answers. Every major AI provider operates with two knowledge layers:

Training data – a snapshot of the internet from months ago, baked into the model's weights. You cannot influence this in the short term. If ChatGPT's training data does not include your brand, no amount of content published today will change what the model already "knows." Training data updates happen on cycles measured in months, not days.

Web search grounding – real-time search results that the AI uses to supplement its training data. When ChatGPT has web search enabled (which it does by default for many queries), it actively searches the web, reads pages, and cites them in its response. This is the channel you can influence immediately.

This distinction matters because most advice about AI visibility conflates the two. "Publish more content" is training-data advice – it might help in 6 months when the model retrains. "Create a substantive pricing page with structured data" is web-search advice – it can get cited within days of publication.

Citaition focuses on web-search-grounded responses because this is the channel where content strategy has immediate, measurable impact.

What Actually Gets Cited by ChatGPT

In our research, we track every citation URL that AI providers use as sources in their responses. This gives us direct evidence of which pages AI is actually reading and referencing when it makes recommendations. Here are the content types that consistently drive citations:

1. Substantive Pricing Pages

Brands with detailed pricing pages – real numbers, tier breakdowns, feature comparison tables, clear explanations of what each plan includes – get cited by AI consistently across all major providers. Gated pricing pages that say "contact sales" receive zero citations. AI cannot recommend what it cannot read.

This is arguably the single highest-leverage page on your website for AI visibility. If you have one page to invest in, make it your pricing page. Read our full analysis of pricing pages and AI visibility.

2. Comparison and Alternatives Content

Pages structured as "[Your Product] vs [Competitor]" or "Best [Category] Tools" are among the most frequently cited content types. When someone asks an AI "what is the best project management tool?" the AI searches for pages that directly answer that question. Comparison content is a direct match.

There is a critical nuance here. When your comparison content gets cited, you tend to win the recommendation. When a competitor's comparison content gets cited, you can disappear from the response entirely. We call this competitive erasure – and in our data, a single competitor comparison post can shape up to 40% of all competitive citations in a category. See how one blog post shaped 40% of competitive citations.

The brands creating comparison content control the narrative. The brands that are not are invisible in the responses that matter most.

3. Deep Product Documentation

Help centres, API documentation, detailed feature guides, and FAQ pages get cited far more than generic marketing pages. The common thread: these pages contain specific, factual, structured information that directly answers questions. AI assistants are optimised to find and extract exactly this kind of content.

4. Third-Party Review Presence

This is where it gets interesting. Different AI providers rely on different third-party sources, and the differences are significant:

ChatGPT's top non-brand citation source is Wikipedia. By a large margin. It treats institutional, encyclopaedic content as the gold standard for authority.

Gemini relies heavily on G2, Capterra, and review aggregators. Structured review data is its preferred authority signal.

Perplexity cites YouTube more than any other source. Video content and the platforms hosting it carry outsized weight.

Google AI Overviews cite industry-specific directories. For SaaS, it is G2 and Capterra. For law firms, Chambers.com and Legal500. For education, CourseReport and CareerKarma. Generic advice about "improving your citations" misses that the sources which matter are entirely industry-dependent.

The practical implication: your AI visibility strategy needs to include third-party platforms, not just your own website. Brand domains account for only 10.5% of Google AI Overview citations. The other 89.5% comes from review sites, YouTube, Reddit, industry directories, and news outlets.

What Does Not Work (Despite What You Have Read)

Our research also identified signals that have weak or negative correlation with AI visibility. These are tactics that many agencies recommend but that our data does not support:

Schema markup alone – having structured data on your pages does not meaningfully improve AI citations. Schema helps search engines understand your content structure, but AI assistants are reading the actual text content, not just the markup.

Trust badges and security seals – these show weak correlation with AI visibility. They may help human conversion rates, but AI assistants do not factor them into citation decisions.

Blog posting frequency – publishing more blog posts does not improve AI visibility. What matters is the depth and substance of individual pages. Ten thin blog posts are worth less than one comprehensive pricing page or one detailed comparison post.

Traditional SEO rankings – ranking well in Google organic results does not guarantee appearing in AI recommendations. The signals that drive Google rankings and the signals that drive AI citations overlap only partially. Many brands that rank well in traditional search are invisible in AI responses, and vice versa.

The Provider Problem: Why a Single Strategy Fails

Perhaps the most important finding from our research is that AI providers disagree on who to recommend nearly half the time. The same query, asked to ChatGPT, Gemini, and Perplexity, produces a different brand recommendation in 48% of cases.

This is not a minor discrepancy. It is structural. Each provider uses different search infrastructure, prefers different source types, and applies different ranking logic. A brand can score well on ChatGPT and poorly on Gemini in the same week for the same queries. See our full provider comparison research.

What this means in practice: optimising only for ChatGPT is optimising for one-fifth of the AI visibility landscape. You also need to account for Gemini, Perplexity, Google AI Overviews, and Google AI Mode – each of which has its own source preferences and citation patterns.

A Practical Action Plan

Based on our research, here is what we recommend – in priority order – for brands that want to appear in AI recommendations:

1. Fix Your Pricing Page

Make it the most detailed, well-structured page on your site. Include real prices, tier breakdowns, feature comparison tables, and explanations of who each tier is for. Aim for 1,000+ words of substantive content. Remove any "contact sales" gating that prevents AI from reading the information.

2. Create Comparison Content

Write your own "[Your Product] vs [Competitor]" pages for your top 3-5 competitors. Be honest and substantive – AI assistants can detect and deprioritise thin promotional content. The goal is to be the authoritative source for comparisons in your category, so that when AI searches for comparison data, it cites your page rather than a competitor's.

3. Build Third-Party Presence

Claim and complete your profiles on G2, Capterra, and any industry-specific directories relevant to your category. Actively collect reviews. These platforms are among the most-cited sources by Gemini and Google AI Overviews – if you are not on them, you are invisible to the AI providers that rely on them most.

4. Deepen Existing Content

Audit your key pages (product pages, feature pages, help documentation) for content depth. AI assistants cite pages that contain specific, detailed, factual information. Thin landing pages with a headline, three bullet points, and a CTA button will never get cited. Invest in making every important page substantive enough that an AI would find it useful as a source.

5. Monitor Across All Providers

Do not rely on manually asking ChatGPT once a month whether it knows your brand. AI responses change with every query, and different providers give different answers. Systematic monitoring across all major AI surfaces is the only way to understand your actual visibility and track whether your changes are working.

The Organic Window

One final consideration. As of early 2026, every AI recommendation is organic. There are no paid placements in ChatGPT, Gemini, Perplexity, or Google AI Overviews. This will not last – OpenAI and others have signalled plans for advertising within AI responses.

When paid placements arrive, the organic visibility landscape will change significantly. Brands that have established strong organic AI visibility before that shift will have baseline data, established citation patterns, and content authority that cannot be replicated quickly. Brands that wait will be competing against both organic leaders and paid advertisers simultaneously.

The window for building organic AI visibility without paid competition is open now. It will not stay open indefinitely.