Amazon shipped six platform-level changes in late 2025 that affect every account we manage. Most operators heard about them in vendor newsletters. Few have actually modeled what they cost or save. Here’s what we’ve seen in practice across the accounts we run.
Project Starfish: AI listing automation you don’t fully control
Amazon’s Project Starfish crawls external brand websites, reportedly 200,000+, and uses that content to auto-generate titles, bullets, and descriptions. It can then run A/B tests on your listings without requiring your input.
The risks are real and underreported. If your brand site uses different specs than your Amazon listing, even slightly, Starfish can introduce inaccuracies at scale. We’ve audited listings where AI-generated bullets misquoted dimensions, dropped warranty language, or stripped out compliance copy that matters in regulated categories.
What we do: audit your external brand presence before Starfish finds it, flag discrepancies between your .com and your ASIN content, and build a review workflow so any AI-pushed listing changes hit a human before they go live.
Seller Assistant: the account health tool with teeth
Amazon’s Seller Assistant now proactively monitors inventory velocity, compliance exposure, and account health metrics, and issues recommendations with deadlines attached. “Recommendation” is doing a lot of work in that sentence. Dismiss them repeatedly and you surface in Amazon’s suppression queue.
The inventory alerts are legitimately useful. Recommendations to mark down slow-moving units before long-term storage fees hit, or to remove stranded inventory before it becomes a compliance flag, save money when acted on quickly. The compliance monitoring is less consistent, we’ve seen it flag products for regulatory issues that didn’t apply to the selling region.
What we do: triage Seller Assistant outputs weekly, separate the actionable from the noise, and implement the high-value recommendations before the escalation window closes.
Amazon DSP + Netflix: what Q4 2025’s media expansion actually changes
Amazon DSP advertisers in the U.S., U.K., Canada, Mexico, and Japan can now buy Netflix inventory directly through the Amazon DSP interface. The pitch is unified planning across Amazon shopper data and Netflix’s audience reach.
The reality: Netflix inventory is upper-funnel awareness, full stop. It doesn’t close purchases on Amazon the way retargeting does. For brands with strong brand awareness goals and budgets above $25K/month in DSP, it’s worth testing. For 7-figure brands allocating most of their DSP budget to retargeting and product targeting, this isn’t a Q1 priority.
We’re running incremental budget tests for interested clients, isolating Netflix spend in a separate line item and measuring branded search lift as the proxy metric, since last-click attribution doesn’t capture cross-screen impact accurately.
Creative Studio AI: faster creative, same review burden
Amazon’s Creative Studio uses AI to generate ad concepts from product data and Amazon shopping signals, cutting production time from weeks to hours for Sponsored Brands and DSP creatives. The output quality has improved significantly from the early beta versions we saw in 2024.
The catch: brand consistency. The AI pulls from your listing content and Amazon’s internal data, not your brand guidelines. We review every AI-generated creative against brand standards before it goes live, and we’ve built prompt templates that reduce the number of revision cycles from four or five down to one or two.
Multi-Channel Fulfillment expansion: one network, more channels
Amazon expanded MCF to include Shopify, Walmart Marketplace, and SHEIN, letting brands fulfill orders from all three channels out of their FBA inventory. The economics make sense at scale: consolidated inventory, one pick-and-pack network, unified reorder points.
The operational complexity is real. MCF orders from non-Amazon channels arrive in Amazon packaging, which some brand operators find inconsistent with their DTC positioning. Stockout risk is also pooled across channels, a viral TikTok moment can drain inventory that was supposed to cover Amazon replenishment. We model shared inventory buffers by channel before connecting MCF to any new platform.
Hispanic market localization: the underused lever
Amazon rolled out Spanish-language ad formats, localized deal placements, and tailored detail page features for the Hispanic market. The addressable audience is large and the competitive density in Spanish-language search is still relatively low in most categories we track.
We’ve seen solid efficiency gains for brands with bilingual product lines who test Spanish-language Sponsored Products campaigns alongside their English campaigns. The keyword list starts thin but builds quickly with search term report data. It’s a low-cost experiment with category-level upside.
How we run these across accounts
None of these tools replace operational judgment. Starfish needs a listing quality gate. Seller Assistant needs triage. Creative Studio needs brand review. MCF needs inventory modeling. The brands getting value from Amazon’s AI rollout aren’t the ones who turned it all on, they’re the ones who built the review layer that makes automation safe to run at scale.
That review layer is most of what we do. If you’re running these tools without it, the risk accumulates quietly until a suppression, a compliance flag, or a stockout makes it visible.