How to get your product recommended by ChatGPT and other LLMs

Executive overview

Most SEO thinking does not transfer cleanly to answer engines. LLMs summarise many citations rather than ranking one blue link — so showing up once at the top is worth far less than being mentioned repeatedly across many sources.

AEO is tractable for early-stage companies immediately: a Reddit comment, YouTube video, or blog mention can surface in ChatGPT answers the same day, with no domain authority required. Traffic quality is higher — Webflow saw 6x the conversion rate from LLM referrals versus Google search.

The core insight: answer engines reward citation volume across many surfaces, not a single high-ranking page, which inverts the logic of traditional SEO.

What AEO is and how it differs from SEO

• AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization) refer to the same thing: showing up in LLM responses.
• LLMs use RAG (Retrieval-Augmented Generation) — they run a search, then summarise the results. This is the layer you can influence; the core trained model is largely out of reach.
• In Google, ranking #1 wins the click. In an LLM, the most-cited source wins the summary slot — frequency of mention across citations matters more than rank position.
• The tail of questions is far longer in chat (~25 words per query average) than in search (~6 words), opening up topics that have never been searchable before.
• Early-stage companies can compete immediately; they don't need domain authority to earn citations.

The citation landscape

• Citations break into groups: your own site, video (YouTube, Vimeo), UGC (Reddit, Quora), tier-one affiliates (Dotdash Meredith: Good Housekeeping, Investopedia, All Recipes), tier-two affiliates, blogs.
• Each group requires a distinct strategy; on-site and off-site work compound together.
• ChatGPT and Google search results overlap only ~35%; Perplexity overlaps ~70% with Google. You cannot assume one strategy covers all surfaces.
• Webflow attributes 8% of signups to LLM traffic — now a top acquisition channel.

On-site strategy

• Create landing pages targeting high-volume keywords; these earn both Google rankings and LLM citations simultaneously.
• Go deep on the long tail: features, integrations, use cases, supported languages — anything a prospect might ask during evaluation.
• Help center content is underused. Move it from a subdomain to a subdirectory, add internal cross-links, and expand it to cover obscure use cases (the "long tail of support questions") that appear in sales calls and customer support tickets.

Off-site and citation strategy

• YouTube and Vimeo: straightforward to produce; B2B topics have very few videos, so even a basic video on a niche topic can dominate.
• Reddit: make a real account, state your name and company, give genuinely useful answers. Five authentic comments can outperform hundreds of fake ones that get banned. The community polices spam effectively, and so does ChatGPT's citation algorithm.
• Affiliates: paying Dotdash Meredith or Forbes to mention your product is expensive but predictable — useful when conversion value is high (e.g. financial products, commerce).
• Paying for affiliate placement requires budget; Reddit and YouTube are low-cost starting points.

Question research and tracking

• Identify target questions by converting paid-search keywords (your own and competitors') into natural-language queries — ChatGPT can do this conversion.
• Mine sales calls, customer support tickets, and Reddit threads for questions that never appear in search data but are clearly being asked.
• Set up an answer tracker (many tools exist, all broadly similar to keyword rank trackers). Measure share of voice: how often you appear across multiple LLMs and question variants.
• Prioritise ChatGPT; also track Perplexity and Gemini given their distinct citation pools.

Running experiments

• Split questions into a control group (no changes) and test groups (one intervention per group: Reddit comments, YouTube video, affiliate placement, etc.).
• Run for several weeks before and after the intervention; compare movement against the control.
• Reproduce results multiple times before treating something as confirmed. Most published AEO best practices are not backed by analysis.

AI-generated content does not work

• Graphite's study across thousands of queries found only 10–12% of Google results and ChatGPT citations are AI-generated; ~90% is human-created.
• AI-assisted content (human in the loop) works and is the clear direction. Fully automated, unedited AI content does not rank.
• The structural reason: search algorithms select for diversity and information gain. A flood of derivative AI content converges on a single point of view ("model collapse"), degrading result quality and triggering algorithmic suppression — exactly what happened with scraped content in 2007.

Key misinformation to ignore

• Google Search is not dying. Google's own data shows publisher traffic is up slightly. New surfaces (TikTok, LLMs) expand the pie; they do not shrink Google's share.
• AEO tools are largely commoditised; some providers charge prices that are not justified by underlying functionality.
• AEO and SEO are more alike than different — same core technology, with meaningful differences only at the head (citation volume over rank) and the tail (longer, more conversational queries).