What’s New in Salsify: AI Agents, Agentic Discovery, and a Bigger Network

Salsify has shipped a dense run of updates this quarter, and the throughline across nearly all of them is the same one we’ve been tracking for a while now: product data has to serve two audiences at once, the human shopper scanning a page and the AI agent summarizing it on their behalf. As we wrote in The End of Pages, The Rise of Answers, visibility used to mean a ranked listing. Now it means being named inside an AI-generated answer, and your product either makes that answer or it doesn’t.

Salsify’s own Q3 2026 product update backs this up with numbers: 60% of consumers are already using AI for online shopping research, AI referral traffic is up nearly 400% year over year, and that traffic converts 42% better than traffic from any other source. Most brands know this shift is happening. Far fewer have operationalized it, with roughly three quarters of organizations naming AI a top priority while only about a third have embedded it into how they actually work.

As a Salsify Platinum Solution Partner, Sitation implements Salsify for clients, runs Salsify-based managed services to keep it optimized day to day, and helps clients stand up and actually adopt newer capabilities like Intelligence Suite rather than letting them sit unused. Here’s what’s new, and how it fits the broader pattern.

Angie Gets a Few Meaningful Upgrades

Angie, Salsify’s in-platform AI agent, picked up several updates aimed at closing the gap between “AI that answers questions” and “AI that does the work.” Text generation tasks now include a live web search toggle, so instead of working from whatever content the underlying AI model last saw during training, Angie can pull a product’s current live listing, say, what’s actually published on a retailer’s site today, before enhancing or rewriting it. That’s a small feature with an outsized purpose: as we’ve noted in our GEO series, keeping data consistent across every retailer is one of the clearest ways to reduce exclusion risk when an AI system is deciding whether your product qualifies for an answer. A tool that can check what’s actually live before touching it is a real assist there.

Angie’s channel mapping skills also expanded from formula generation into direct property mapping and single value mapping, moving her from a narrow mapping-formula assistant into something closer to a full channel mapping assistant. Formula help, previously scoped to channel mapping, now extends natively to computed properties as well. Salsify’s roadmap points toward a bigger shift by early 2026: rather than just answering a question or helping build something, Angie is expected to move toward taking the action herself, applying a filter or building a list directly from a plain-English request.

Image Generation Gets More Control

Salsify’s AI image generation task, which lets teams generate new product visuals directly inside the platform rather than routing every request through a design queue, picked up two new controls this release: a dropdown for image type and size (landscape, lifestyle, and other formats) and a choice of output file format, currently PNG or JPEG, with a custom dimension option for anything outside the standard presets. One current constraint worth knowing: the image generation task only works with Salsify’s own AI model for now, not the other AI providers available elsewhere in the platform.

Closing the Loop From Content to Discovery

We’ve argued before that brands need to think in terms of two distinct shelves: the open-web AI systems that build influence, and the retail AI systems that actually decide which products get recommended at the moment of purchase. Salsify’s Site Catalogs update speaks directly to that second shelf. Catalogs already had a toggle to optimize for agentic search; catalogs with that toggle enabled now show a new Bot Traffic panel under Site Analytics, showing how many bots are crawling a given catalog and, where identifiable, which ones. It’s a meaningful addition for any team that’s invested in making content agent-readable but has had no way to confirm whether agents are actually reading it, which connects to a pattern we described in Win the Shelf When Competitors Go Dark: out-of-stock moments are increasingly recommendation moments, and knowing whether an AI agent is even looking at your catalog is a prerequisite to winning that moment.

Worth flagging alongside this: classic catalogs are officially sunsetting December 31, 2026. Anyone still running classic catalogs should treat that migration as a near-term project, since the discovery features described above are exclusive to the newer Catalog Sites.

Toward Touchless Syndication

The most significant channel-side update is Auto Map, an AI-assisted channel mapping tool currently in a limited pilot, with a broader rollout expected in phases starting in October. Rather than mapping every attribute manually when setting up a new channel, Auto Map suggests mappings based on patterns from how similar attributes are typically mapped across comparable channels. It’s currently capped at 250 attributes per channel, and Salsify has been upfront that mapping accuracy is still being refined, human review of suggested mappings remains part of the process. It’s an early-stage feature, but it’s a concrete step toward the touchless syndication vision Salsify has described for a while, alongside a longer-term Autofix capability expected in early 2027 that would automatically surface and suggest fixes when a retailer changes its requirements.

On the GDSN side, product data validation is shifting from reactive to proactive: per-product validation ahead of sync is already available, with bulk validation across multiple products at once on the near-term roadmap. Later this year, Salsify is also planning to become a certified GDSN data recipient, letting customers automatically subscribe to and ingest GDSN data directly in the platform.

A Bigger, Smarter Network

Several retailer and distributor connections either went live or expanded this quarter. Schnucks, a grocery retailer Salsify has connected to for some time, moved from using brand-published content for internal purposes only to actually displaying that content live on its consumer site as of August, with enhanced content support following shortly after. Quill, a B2B subsidiary of Staples, is a new enhanced content destination, with a full channel connection expected to follow. And Johnson Brothers, an alcohol and beverage distributor, is a new channel connection, notable both because it’s Salsify’s first connection with this type of distributor and because it reuses most of the same attribute set as Salsify’s Open Catalog.

On the Amazon side, Amazon Shop Direct (formerly Shop the Web) lets brands get product data in front of Amazon shoppers and route them straight to the brand’s own site, without managing a separate Amazon listing or seller account, and without Amazon charging any fee to participate. Separately, Amazon A+ Premium continues to add modules built specifically for AI discovery, structured Q&A, comparison tables, and rich text that AI models can parse directly to answer more complex shopper questions. That’s consistent with what we’ve observed about how Rufus specifically evaluates a listing: it prioritizes context-rich, problem-solving copy and complete attributes, and returns only a handful of products per query. A brand’s data either supports that kind of question or it gets excluded from the answer entirely.

What Proven Adoption Actually Looks Like

The most useful data point from this quarter isn’t a feature at all. Kevin Creese, director of ecommerce at Wastequip, described on Salsify’s Q3 webinar how his team used Angie and Intelligence Suite to standardize naming and categorization across roughly 60,000 SKUs, building in human review steps that got progressively lighter as accuracy climbed from around 70% to more than 95%. The result, by his account: maintenance time cut roughly in half, and organic and direct traffic now driving more than 60% of visits to the company’s B2B site.

That’s a real-world version of the structure workstream we’ve written about in our own GEO strategy series: consistent naming and complete attributes aren’t a hygiene task, they’re a precondition for being included in an answer at all, whether that answer comes from a shopper’s search or an AI agent’s summary. Wastequip’s traffic and time-savings numbers are what that precondition looks like when it’s actually in place.

That outcome didn’t come from flipping a switch. It came from a deliberate rollout: starting with a narrow, well-scoped use case, building in review steps, refining prompts through several rounds of testing, and only scaling as confidence grew. That’s the kind of work Sitation does for clients standing up Intelligence Suite, scoping the use case, building on proven workflow patterns, and staying involved through managed services once it’s live.

If you’re weighing what any of this means for your own Salsify instance, whether that’s the classic catalog sunset, piloting Auto Map, or finally putting Intelligence Suite to work, we’re happy to talk through it.