It’s always nice for a marketer to get a compliment on their website. But does the compliment still resonate when it comes from an AI agent?
Ask again in a few months and we might have an answer.
On Tuesday, customer journey intelligence platform Knotch declared a new solution called Ace for Agents. The product builds on its existing AI marketing solution, Ace, but is specifically focused on making website content more accessible to AI agents.
But what does it mean for content to be accessible to agents that, by definition, can access pretty much all content across the web?
Effectively, it involves reformatting content into simpler terms and unifying it from multiple sources onto a single webpage, noted Andrew Bolton, chief customer officer at Knotch.
Sources indicate that most LLM crawlers declare themselves when they access a webpage, so Knotch knows to show them the AI version of a page, mentioned Bolton. For those that don’t declare, Ace can analyze actions like on-site behavior to determine that the viewer is a bot.
Industry observers note that AI agents, unlike humans, don’t care about interactive features or aesthetically pleasing graphics. Instead, their focus is on brevity and direct statements.
According to recent reports, often, the reformatting includes adding FAQs and concise summaries, as well as removing JavaScript. The result is a pared-down version of the webpage a human would see, but the key content remains.
According to recent reports, that’s what Bolton describes as a one-to-one version of an existing webpage, in which the information that’s already on it is simply restructured, but nothing new is added. That’s one of the two types of pages that Ace for Agents can develop – the simpler, “v1” offering, said Bolton.
The slightly more advanced version of Ace for Agents involves “injecting newly introduced components of information” from other sources of the brand’s choosing, he mentioned, like third-party review sites, or the brand’s own YouTube channel.
For instance, he added, if a car manufacturer had a particularly favorable review on a third-party site comparing one of its models to a competitor’s model, Knotch can ingest that into the manufacturer’s context model.
It has been revealed that reviews “would be valuable to a consumer who’s trying to evaluate those two different vehicles,” he added. Marketers have a say on whether they want to manually flag external reviews and other content to pull in, or if they want to use an agent to automatically pull in upcoming, positive content, which a human can then review.
The resulting webpage is a “super piece of content” that is able to “combine the atoms” of several different pages across the web, mentioned Bolton. Now, an LLM only needs to access one page, but gets the same amount of data that it historically would’ve collected from crawling all of the data sources individually.
It’s worth noting that the data inputs are entirely curated by the brand. In other words, the good reviews will be ingested, and the bad ones won’t. A brand wouldn’t elect to put negative reviews on their website, Bolton remarked, and Knotch is effectively creating an agent-readable webpage that pulls all of the informative and positive information on the brand into a single source.
As marketing and the open web rapidly evolve, agencies need to build an “experience that reaches both the people and the machines,” Melissa Berger, chief solutions officer at Digitas, told AdExchanger. “Earning attention” has always been a major focus for the agency, Berger remarked, “and now that’s not just earning from the human; it’s earning from the agents, too.”.
Digitas has been working with Knotch for over a year and began testing Ace for Agents several months ago as a way to earn some of that agentic attention.
In a recent development, although Ace for Agents has been a good way for brands to increase discoverability, they still need to understand who their human customers are, mentioned Berger. Otherwise, the LLM-accessible site may be full of information, but not have the answers to the specific questions the humans are looking for.
Once a brand understands exactly what sort of questions its customers are asking and what queries it should aim to show up against, mentioned Berger, “the rest of the content is updated to be catered more toward that.”.
