How AI shaped holiday shopping and what it means for content in 2026

What changed during holiday 2025

This year witnessed the most significant change in shopping behavior since the rise of e-commerce. Traffic from AI assistants, like ChatGPT and Gemini, grew 119% year-over-year, converting at 700% higher rates than social media, according to Salesforce. By Cyber Week ‘25, Salesforce reported AI and agents influenced 20% of all global orders, driving $67 billion in sales.

While shoppers broadly adopted AI tools this year, many brands didn’t. With cost pressures and competition mounting, 2026 is the window to build content operations that power web, mobile, and agents from one structured foundation.

The content operations trap

When a major home audio company built their AI shopping agent this year, they created a separate content team that emailed PDFs of product specs to developers. Manually. One product at a time.

They were already running their product catalog and support articles in Sanity, which fed to their website. They just didn't see the connection. To them, the agent was a new product that needed new content infrastructure.

This is the same mistake businesses made 10 years ago managing website content in files instead of systems that led to slow, duplicative, and inconsistent content. The difference now is the technology exists to structure content once and deliver it to any channel.

Simply put, agents are just another channel. The content model powering your website, mobile app, and channels should power your agents too.

Same source of truth, same workflows, same team managing it.

With Sanity, your content operations already exist. The infrastructure is ready to adapt to whatever comes next.

What agents need from your content

Sanity's data-first approach to content unlocks capabilities that traditional CMSes can't offer. Next we'll cover what this foundation enables for AI.

1. Getting LLMs, like ChatGPT, to find and recommend your products.
2. Running AI content operations with Sanity’s Content Agent to create, translate, and publish content faster.
3. Using Markdown for AI-ready support docs
4. Building your own on-site agent that guides customers through purchase considerations.

Contextually-rich LLM discoverability

Shoppers are increasingly asking AI for hyper-personalized recommendations. Not “best wireless headphones,” but “wireless headphones for my sporty friend, who dislikes in-ear styles and listens to bass-heavy music.”

These LLMs serve as a real-time concierge for customers, but they process information differently than humans. Where humans skim headlines and images, agents consume every word on the page – specs, compatibility notes, FAQs, support docs, everything. That means content architecture matters more, not less.

Making technical docs AI-ready

Markdown, a lightweight-plain-text format, consumes fewer tokens than HTML and provides clear structure that LLMs can easily parse. It’s highly effective for passing information to LLMs, reducing API costs while improving comprehension. LLMs understand its syntax and can generate responses in the same format, making it easy for both humans and agents to read and edit.

For our own docs, Sanity converted its documentation to .md files. Here is an example of our MCP Server documentation and its .md version.

Content teams need to move faster

Content can’t be static. When your product catalog syncs with your content system, spec updates flow automatically while preserving the brand storytelling that leads to more purchases.

Sanity’s Content Agent and AI capabilities understand your schema, something no other CMS offers. It can scan your entire content library to surface outdated claims, identify gaps, update metadata, and translate for localization. Because it understands your content model and business logic, AI works within your workflows and governance, not as generic plugins bolted on after.

“I asked Content Agent to review a promotion, and it immediately spotted that our legal dates didn’t match the start and end dates in the content. This is exactly what I need.” - Kenny Ashton, Senior Product Manager, RONA

Building your AI shopping agent

75% of retailers say AI agents will be essential to gain a competitive edge by 2026, according to Salesforce. The immediate use case: on-site shopping agents that guide customers through complex product decisions, such as compatibility checks, budget, and preferences, with the same conversational ease they get from ChatGPT.

If your content is already structured as data, and synced with your product catalog in a single source of truth, you’re ahead. AI agents can plug directly into existing content and workflows.

No new systems, no blocked developers, no separate content operations.

What to do before 2026 shopping season

You have roughly 10 months before Holiday 2026. Here's where to start:

1. Test your products in AI tools right now. Query ChatGPT, Gemini, Claude, etc. about your key products. Document what gets recommended and what doesn't. Identify content gaps where they have outdated or wrong information.
2. Ensure your pages are server-rendered. AI crawlers don't run JavaScript. Sanity works well with server-rendered architectures, so your product content is visible to AI-powered search from the start.
3. Make content easy to use in AI tools. Customers and sales teams increasingly paste product info into AI assistants. Sanity's structured content can be transformed to Markdown on demand, and you can add "Copy for AI" features without maintaining separate content.
4. Archive or update old content. LLMs often surface old content more than search because it’s looking for relevant information. Set up a content review cadence, especially for product launches, FAQs, and community content.
5. Verify that your CMS is AI-ready. Is your product content already structured as data? Can you query it programmatically? Assess if your CMS is ready and future-proof for AI.
6. If you’re already using Sanity, talk to your account manager or contact sales about how Sanity’s Content Agent can help your team today.

The 2026 competitive divide

2025 was the observation year, where AI shopping went mainstream.

2026 is the execution year. AI discoverability and agent commerce become table stakes.

2027 is too late. The competitive advantage has already been established.

By Holiday 2026, every major e-commerce site will have some form of AI shopping assistance. The winners won’t be those with the flashiest chatbot, they’ll be those who built a content foundation in Q1 and Q2 ‘26 that powers web, mobile, and agents from a single source.

The laggards will still be manually syncing product specs and localizing content, fixing emergency bugs when agents recommend wrong products, and building separate AI content teams that duplicate work. Taking weeks to ship new content and features, while the winners publish and sell out products in minutes.

The foundation gets built in quarters, not sprints.

The window is closing.