Profound's James Cadwallader: brands are about to lose most of their consumer relationships to AI answer engines

Summary

James Cadwallader, co-founder and CEO of Profound, is building what may become the default infrastructure layer for brand visibility in an AI-first world. The New York-based startup, launched in August 2024, raised a $3.5 million seed round co-led by Coastal Ventures and Saga, and has moved quickly into enterprise accounts including MongoDB, Indeed, Mercury, Ripling, and Ramp.

The core argument is structural and urgent. Traditional search connected users to brand websites; AI answer engines sit in the middle and absorb the relationship. When a consumer asks ChatGPT about a product, follows up with more questions, and eventually transacts inside the same interface, the brand never touches that journey. Cadwallader says most brands haven't fully grasped this yet, but they are about to lose the majority of their consumer relationships to answer engines. A zero-click future, where users browse, research, and buy without ever visiting a brand's site, is the direction OpenAI and others are heading.

What Profound actually does

The platform runs large-scale interrogations across AI answer engines, sending thousands of open-ended, non-brand-direct queries to see which sources ChatGPT, Perplexity, and Copilot pull from when answering questions in a given category. That source mapping becomes the basis for visibility strategy. Ramp used the platform to increase its visibility 7x in one tracked index. A separate unnamed enterprise client with 12,000-plus employees ran a global PR push around a product launch and saw a 50% increase in AI answer visibility.

On the tactical side, Cadwallader points to a few things that are working: structured data over verbose prose, because the audience is a retrieval agent rather than a human; metadata that summarises the full article, which engines often pick up directly; and llm.txt files, which Profound's own site is now testing. Profound's llm.txt is getting as much traffic as its sitemap. For content-heavy sites like Stripe, an llm-full.txt that consolidates all documentation into a single text file may prove especially effective, since answer engines handle large text inputs without the friction they have with multi-page crawls. Images, by contrast, are largely irrelevant to retrieval.

Incumbents are more aware than the early-2000s SEO catch-up narrative suggests. Large Fortune 100 companies are already using Profound, though they won't allow logo disclosure. The lag isn't awareness — it's the gap between understanding the problem and building internal workflows to act on it.

Ads inside LLMs

Cadwallader sees generative advertising in conversational interfaces as potentially one of the most powerful unlocks in marketing history. A language model that can synthesise a perfectly tailored ad for the exact moment and context of a conversation is a fundamentally different medium from display or search. He expects the market to be more fragmented than Google search became — partly because more capital is explicitly trying to prevent another monopoly — but he's cautious about timelines, particularly for OpenAI, where he thinks the corporate structure creates constraints on ad-model development.

Profound's near-term bet is measurement and optimisation. The paid-ads layer, if it arrives, is an obvious extension of that position.