👋 Hi, it’s Kyle Poyar and welcome to Growth Unhinged, my weekly newsletter exploring the hidden playbooks behind the fastest-growing startups.
There’s no shortage of sketchy advice about how to hack ChatGPT, Perplexity, and Google Gemini. But here’s what matters: these tools are becoming the most influential gatekeepers in the B2B buying journey, and they play by different rules than traditional search.
ChatGPT alone now has a billion active users and was the fifth most-visited website in the world in June 2025. Usage is up to 2.5 billion prompts per day, which is quickly catching up to Google’s 14 billion daily searches. Meanwhile, Google is steadily morphing into an AI-powered answer engine itself, with AI Overviews appearing in 13% of all searches (up 2x since January).
Increasingly, buyers aren’t researching products themselves. They’re asking AI what to buy, who to trust, and how to decide. That means if your product isn’t mentioned in these answers, it may not exist at all in the buyer’s mind. And if you are mentioned, AI search could become your fastest-growing source of pipeline. As my friend Sam Richard, CRO at ngrok, recently told me, “It’s driving a truly shocking amount of business for us.”
The goal is no longer getting clicks to your website, although that wouldn’t be such a bad thing! It’s about being mentioned by AI answer engines in a way that influences the right customers (and does eventually lead them to your website). And doing that requires an answer engine optimization (AEO) strategy. (Folks can’t seem to agree on the name of this, by the way. I’m calling it AEO, but have also seen it called AI SEO, GEO, LLMO, AIO… LMAO).
To learn how to get found in AI search, I turned to Josh Blyskal who leads AEO strategy and research for Profound, one of the emerging breakout players in AI search. I’ve summarized the conversation into nine practical takeaways that you could start applying now to improve your AI search visibility:
Decide which prompts you want to own
Be aware that answer engines behave differently
Write specialized, specific content
Create surround sound around your product
Get mentioned on Reddit
Use structured comparisons and lists
Make fewer images and more tables
Measure your influence, not just clicks
Stay on top of citation drift
Decide which prompts you want to own
Answer engines are actively recommending products (and newsletters 😉). Similar to SEO, you need to decide which prompts are most critical for your brand based on two factors: the volume of queries and the relevance to your product and its core differentiators. It’s worth mentioning that tools like Profound’s Conversational Explorer can be helpful here.
In my case, this might be queries like best newsletters for startup founders or best newsletters for startup pricing. For a shoe brand, it might be best running shoes or most comfortable shoes or whatever is most important for them to dominate.
Josh recommends including three types of prompts in your target list:
Core prompts that are must-win (~33%)
Competitive knife-fights where you’re trying to be found, but aren’t the true cornerstones of your strategy (~33%)
Experimental prompts where you perhaps haven’t focused historically, but could leapfrog the competition (~33%)
Be aware that answer engines behave differently
Josh compared 100,000 prompts across ChatGPT and Perplexity to see how much they overlap. Turns out, not by much. Only 11% of domain citations were shared between the platforms.
It is possible to be dominant in ChatGPT, but invisible in Perplexity. Start your answer engine optimization (AEO) with the most important platform for your audience (likely ChatGPT or Google AI Overviews) and build strategies specifically for that platform. Then expand your surface area over time.
Write specialized, specific content
To get mentioned in these prompts, you’ll need to build out new content. Your goal: isolate a topic that everyone’s talking about, then write the most definitive and specific content around it.
If you’re operating a shoe brand, you’d want to write hyper-specific content like why cushioned running shoes are better for your knee health. And then you’d want to cover 20 different options for cushioned running shoes along with a hyper-specific comparison for each.
“It’s a race to granularity,” Josh said. “Hyper-specificity is key. You want to own your product niches and differentiators.”
Create surround sound around your product
Similar to backlinks in SEO, you want your product and content to get amplified by the voices that AI search engines trust. The best starting points are user-generated content and social media sites like Reddit, LinkedIn, YouTube, and Quora; these all rank highly among the top cited domains in AI search.
Josh believes there are some interesting arbitrage opportunities here right now. Niche blogs or niche creators could have a small following, but a surprising share of answer engine citations. These can make for great affiliate opportunities.
Get mentioned on Reddit
The single best source of surround sound: Reddit. It’s now the most-cited domain within ChatGPT and Perplexity, accounting for about 6% of their citations.
Josh recommends getting your name in the right Reddit threads by any means necessary, whether as the brand or anonymously. Even a comment like Profound is great can make a big difference.
There are more and less risky ways to do this. A starting point is to create a Reddit user page, which gets your information on the Reddit domain. A much touchier approach would be to comment on specific threads that tie back to your target prompts. If you’re thinking of doing this, make sure the comments don’t only reference your own product; also include a few sentences about competitors.
Use structured comparisons and lists
AI answer engines are essentially outsourcing their thinking to others who do the research and collating for them. These answer engines like to grab things in chunks, which means they prefer highly structured content, punchy paragraphs, and distinct lists.
One-in-three AI answer engine citations are from comparative listicles, Josh tells me. That’s the single best performing type of content (after that it’s blog content, representing one-in-ten citations). When in doubt, use a comparative list.
Make fewer images and more tables
These answer engines can’t see images well (at least for now), Josh highlighted. They prefer structured tables built-into your pages.
Charts or images used to be the main way of showing information (y’all know I’m a fan, especially because they’re great for LinkedIn). But if you want to be cited by ChatGPT, you’ll want to reduce the number of images in your posts. When you do have images, make sure to add really specific alt text that describes what’s in the image. Or turn the image into a table.
Measure your influence, not just clicks
Many brands are seeing referral traffic from ChatGPT or Perplexity sky-rocket. But the reality is that the vast majority of AI search is zero-click. Someone gets an answer from AI, then moves on. Clicks and website traffic aren’t going to be the best way to measure how you’re doing.
This means going back to brand-related KPIs like share-of-voice, i.e. how much your product is being talked about relative to your main competitors. I’d also encourage folks to consider manual attribution methods like adding a “How did you hear about us?” field or question to a lead form. Realistically, manual attribution will also help uncover other forms of non-click influence as well, including podcasts, word of mouth, and community efforts.
Stay on top of citation drift
A final word of caution: citations from answer engines are seeing a ton of drift. Josh analyzed 80,000 prompts across answer engines and found that almost half of the domains cited shifted within a single month.
My expectation is that this drift will continue as LLMs keep evolving and figuring out the most relevant content for a given prompt. I suspect these citations will become even more personalized, too, as answer engines learn from their prior context.
In the short term, you’ll want to refresh your prompt analytics fairly frequently (weekly is a good starting point). And be ready to adapt as LLM preferences keep shifting.
The TL;DR:
Increasingly, people aren’t doing their own research to find your product. They’re turning to answer engines to do it for them. Which means those answer engines get to decide if you’re in the consideration set.
That’s a massive shift in how discovery works. And creates a big opportunity for whoever moves first. Hopefully this is the starting point for your next best growth channel.
Loads of junk advice on this topic online right now. But this article is the real deal. Thanks Kyle!
Just this past week, I've learned how important llms.txt and llms-full.txt are for AEO/GEO. I thought it was a niche topic until I learned that:
a) 90% of B2B buyers are using ChatGPT and similar tools. If you're not doing AEO, you're invisible.
b) Anthropic (which knows a thing or two about LLMs) now has an llms.txt and llms-full.txt; marketers would do well to follow their example to succeed in the nacent world of AEO.
Here are their files:
https://docs.anthropic.com/llms.txt (basically a markdown version of their doc sitemap)
https://docs.anthropic.com/llms-full.txt (markdown version of all their documentation)
There's a lot more to explain here, which is why I put together this video in case it's of interest: https://www.youtube.com/watch?v=3ERlcDoKqlE