.avif)
Table of Contents
Key takeaways
- Answer engine optimization (AEO) is the practice of ensuring AI systems can retrieve, understand, and accurately cite your brand. It involves auditing how your company appears across large language models (LLMs), identifying gaps in coverage, and shaping the content and structure those systems rely on.
- Early ChatGPT couldn’t reliably replace search because it wasn’t connected to the live web. Once answer engines started synthesizing from real-time sources and citing them, AEO became an inevitable strategy for brands.
- In the future, generative AI will move beyond chatbot interfaces and into “generative UI,” where the model assembles a real-time experience with the components you need to explore options, compare products, and complete actions on the spot.
- For many brands, the most direct lever is still publishing content on their own website. The goal is to give AI enough information to accurately answer real questions about your business. It's also to structure that content so that it can be easily understood by bots.
- FAQs are one of the highest-leverage formats for AI search. They contain both the question and the answer, which makes them particularly useful for AI models. In practice, teams have seen upwards of a 20% lift in page citation when they append FAQs to key pages.
- As humans visit websites less frequently, ROI must be measured more creatively. Citations and user-agent fetches are the real-time signals left when an AI system retrieves your content to answer a question. Those network logs act as a proxy for human interest.
- Simple attribution still works. Several teams have unlocked measurable pipeline growth just by adding “ChatGPT” or “AI search” as an option in their “How did you hear about us?” form.
For years, digital marketing has treated visibility as a traffic equation, with search rankings driving clicks and clicks driving conversions. But as tools like ChatGPT, Gemini, and Perplexity become the first stop for information retrieval, that model is starting to break down.
Profound is a platform that helps companies understand how they show up in AI search. Since launching the company, founder and CEO James Cadwallader has worked with marketing teams navigating the shift from links and clicks to citations and synthesized answers, including 10% of Fortune 500 companies.
We sat down with James to unpack what this shift really means – why web-connected AI made AEO inevitable, how generative UI will change the search experience, and what brands should be doing right now to stay visible. This conversation is part of Pigment’s Perspectives series, where Pigment's CEO Eléonore Crespo speaks with business leaders shaping the way modern companies operate.
Information retrieval is having its CDs-to-streaming moment
James believes AI answers are on their way to becoming the default interface for retrieving information.
The analogy captures the directional nature of this shift. While links won’t disappear overnight, the default behavior of internet users is changing. Once the status quo has been upset, the ecosystem reorganizes around it.
AI hasn’t always been as reliable as it is today
At the beginning of 2024, ChatGPT had a big problem: It wasn't connected to real-time web data. Whenever you asked it a question, the answer would consider information up to the cutoff year of its training data. Because of this, every output was a point-in-time snapshot of what we used to know.
This technology wasn’t enough to replace web search, because people want access to information that’s true right now. But, later that same year, Perplexity proved that it was possible to deliver real-time answers sourced from the web.
Imagine communicating with AI through an interactive web page, generated in real time
If the first shift is from links to answers, the second is from static pages to dynamic, generated interfaces.
That page won't stay static for long. He expects responses to evolve into images, video, and interactive components, all spun up in real time so users can navigate, compare, and transact without leaving the interface.
James describes this next layer as “generative UI.” He offers a visual exercise to explain the concept.
With generative AI companies like ChatGPT, Claude, and Perplexity eager to keep users within their conversational interfaces, earning clicks from these sources is becoming less important. Instead, your primary goal with AEO should be to make sure your product, positioning, and competitive context are accurately represented inside a chat experience you don't control.
The lowest-hanging fruit is to publish more content on your website
When asked for practical steps, James points to what he considers the most accessible lever: creating and publishing content on your website. It might seem unglamorous, or obvious even, but it's what the internet runs on.
Content supplies the raw material that answer engines rely on. So brands need to focus on covering real buyer questions thoroughly and accurately.
Answer engines need structured, explicit explanations that remove ambiguity. If a model cannot clearly retrieve and cite your information, it cannot represent you well.
This means expanding beyond traditional top-of-funnel content into content that’s aimed at machines. Maybe they’re not all linked in your homepage header, but they exist so bots can get the information they need in real time.
There are other tactics also worth exploring, like Reddit, traditional PR, and YouTube content, which is especially prevalent inside Gemini results. In most cases, you should consider these only after you’ve prioritized the content on your website.
Adding FAQs is a practical tactic that works right now
Answer engines process billions of prompts, including highly specific, high-intent queries that rarely justify a polished blog post.
FAQs translate buyer intent into clear, structured content that increases the likelihood of your page being cited by LLMs. Structurally, FAQs contain both the question and the answer in explicit language, making them easy for models to retrieve, interpret, and reuse.
Instead of forcing an AI system to infer intent, you are doing the legwork for it.
For teams looking for a contained experiment, adding high-quality FAQs to key evaluation pages is a pragmatic starting point. A SaaS business, for example, should aim to answer questions like whether a product integrates with a particular stack, how pricing scales across usage tiers, what security controls exist for a given workflow, and how long implementation takes.
So, how do you measure ROI when clicks don’t tell the whole story?
While you can track human referral traffic from AI sources, those numbers will be slightly skewed because many users don't click through to web pages. When attributing value in AEO, the closest thing to clicks is citations.
James says that, in some cases, ChatGPT alone can account for upwards of 10% of referral traffic.
Some of the most practical ways to attribute value to AI search are also the simplest. For several of Profound's customers, adding "ChatGPT" or "AI search" as an option in their "How did you hear about us?" form has enabled them to attribute hundreds of thousands of dollars in pipeline. While it doesn't solve measurement entirely, this simple tactic can surface demand that would otherwise remain invisible.
To keep AI from hallucinating about your brand, make updates often
LLM hallucinations are reducing by the day, but they still happen. An AI system might surface outdated pricing, misattribute a feature, or describe your product based on information that was correct at some point in time because someone published it.
The most common culprit is stale content. An old blog post, an outdated partner page, or a third-party review with last year's pricing can quietly shape how AI answers questions about your brand.
James shared an example from a financial institution that was being described incorrectly in AI answers. The information the LLM was sharing wasn’t fabricated; it had been accurate at the time when it was first published in a third-party article. But the article hadn’t been updated in a while, and that outdated page was shaping how the brand appeared in LLM responses. Once the team identified and corrected the source, the incorrect mentions disappeared.
In practice, that means auditing third-party sources that mention your company and asking them to update what’s no longer accurate. It also means refreshing your own pages regularly. And it means actively monitoring how your brand appears across major LLMs, so you can catch gaps before they become the answer a buyer sees first.
How Profound uses AI internally
When it comes to AI adoption inside the company, James indexes on people with high agency and curiosity – people who want to tinker. He sees fluency with AI tools as table stakes, similar to saying you use Google Sheets.
One of the tools he recommends from their stack is Dust AI, an internal knowledge base that pulls together data sources and plugs them into Slack so information is accessible in real time. With the right tools and the right people, AI adoption becomes less about developing rollout plans and more about removing friction that gets in the way of speed.
Profound's growth has been fast. The company now works with 10% of the Fortune 500, and its team spans offices in New York, San Francisco, London, and Buenos Aires. But James doesn't attribute that momentum to a hiring playbook or a rigid organizational model.
He describes building the team as a more intuitive, values-driven process. One of the most important pieces of advice he received early on came from investor Keith Rabois: don't follow anyone else's rules. Assemble the leadership team that fits the company you're actually building.
James’s own leadership style emphasizes decision velocity over operational playbooks. His belief is that smart, ambitious people can figure out most things, including how to manage teams and set up business processes when the time comes.
Conclusion
AI answers are becoming the default layer for information retrieval, including information about your business. A buyer may never visit your website and still form a strong opinion about your company just by using LLMs.
The playbook James lays out isn't complicated. Publish more content, structure it for machines, and use FAQs strategically to reduce ambiguity. It also involves expanding measurement beyond clicks to include citations, agent visits, and attribution signals that reflect how discovery actually works today.
Inside Profound, he applies the same mindset to building the company. He hires for curiosity and agency, keeps process light, and treats AI fluency as table stakes.
That intensity fuels a larger ambition. James doesn’t see Profound as a point solution for AI search. He sees it evolving into a workbench for marketers – a system of record for how brands are showing up.
.webp)
