PepsiCo’s sustainability team has refocused its annual sustainability reporting to meet the needs of the AI bots that are becoming increasingly important consumers of online information.
Many companies publish lengthy sustainability reports and accompanying data tables, usually as PDFs. But the bots that crawl the internet for information to add to ChatGPT, Claude and other AI models find PDFs more challenging than webpages to navigate.
To accommodate this new class of readers, the PepsiCo team made a series of tweaks to its publishing strategy.
Webpages over PDFs
The food and beverage company published a 2025 ESG summary in PDF form, but at 20 pages the document is less than half the length of the 2024 version. Some of the information that used to appear in the document can now be found on PepsiCo’s ESG Topics A-Z webpage, which includes sections on agriculture, environmental impact and other issues.
Regular updates, not “an avalanche”
“What we’ve been doing from a reporting or communications perspective is we hold all the information and then we put it all out once,” said Dan Strechay, a senior director on PepsiCo’s Global Corporate Affairs team. “Then it’s a huge avalanche of information,” he added — some of which inevitably gets lost.
PepsiCo now makes regular updates to its A-Z pages instead, sometimes in anticipation of heightened interest and additional AI queries on a particular topic. The A-Z page on water use, for example, was updated with the latest data ahead of this month’s World Water Week.
“We take a modular approach in partnership with our reporting team and our legal and control colleagues,” said Strechay. “We say, ‘As soon as the information’s ready let’s put it out.’”
AI-friendly page structures
PepsiCo’s A-Z pages make careful use of section headings, bullet points and other formatting that creates a clear structure.
“Clearly delineating subsection headers and consistent sections — you know, it sounds kind of mundane,” said Anna Palazij, the company’s vice president for sustainability. But, she added, it makes a difference in how AI is able to parse the page.
The pages are also timestamped so that readers — humans and machines — can immediately see how fresh the information is.
What success looks like
Palazij and Strechay still want human readers to visit PepsiCo’s webpages and download the company’s reports. But in an era of AI services that are designed to directly answer questions rather than redirect users elsewhere, as traditional search engines do, they say it’s more important that users get the correct information — be that on a company webpage or via an AI chatbot.
“If it’s coming through correctly on Claude, that’s fine,” said Palazij. “But if it’s not, then that could actually be a detriment. Which is one of the reasons why we tried to make the data accessible by AI.”
