Changelog

When an AI engine answers one of your prompts, it quietly runs its own web searches first, then writes the answer from what it finds. Until now those searches were invisible, so you were optimizing for the answer without knowing what the model actually looked for to build it.

Query Fanouts is a new page that shows you exactly those searches. See the number of distinct fanout queries, filter them by type (Search, Shopping, Synthetic), and group them by topic or prompt. It surfaces which brands the engines name most often inside their searches and the word sequences that come up again and again, so you know what the models look for before they write. Find it in the sidebar.

AI Shopping in the MCP

AI Shopping is now in the MCP and Customer API. Ask Claude how one of your products performs in ChatGPT shopping and to rewrite its product page to close the gaps, or pull product visibility, win rate, ratings, and competitors into your own scripts. Alongside that, the star rating ChatGPT shows next to each product is now visible in Peec, the products table filters by brand, and the product detail view shows the shopping and web searches that surfaced each product.

Filter every chat by what it used

The all-chats view now filters by what a chat actually did: Web Search, Shopping, Product Comparison, Ads, and Map, per chat or aggregated per prompt, so you can answer questions like "which of my prompts trigger Shopping." Chat brand filters also gained AND/OR logic, with a "no brand mentioned" option to surface the answers where nobody in your set came up.

Branding and intent tags

Every prompt is now classified automatically: branded vs non-branded, and by intent (informational, transactional, commercial). Both are available as filters, so slicing your visibility by buyer intent takes two clicks instead of a manual tagging pass.

Brand-mention filters on Sources

The Domains and URLs tables now filter by which brands are mentioned, with the same AND/OR logic as chats. It is the quickest way to find pages that talk about you and a competitor together, or pages citing a rival but not you.

A redesigned members view

Managing access is now one company-wide members page with role-based access, plus search and CSV export. If you run many seats across projects, seeing who has access to what is a lot easier.

Shoppers now ask ChatGPT what to buy and trust the answer more than any search result or paid listing. You can see your Google rankings and your conversion rate, but when a shopper asks an AI what to buy, you have no idea which of your products it recommends, where it sends the buyer, or whether your visibility is growing or quietly slipping. For most e-commerce brands, AI commerce is a blind spot.

AI Shopping Analytics closes it. Bring your catalog and Peec gives you SKU-level visibility into AI shopping: which products get picked, why they get picked, and where AI sends your buyers. Today it tracks ChatGPT, where the shopping experience is most advanced, with other engines to follow. Go to Shopping in the sidebar to connect your catalog.

Every product scored

Your catalog becomes a table, one product per row, each with the metrics that decide whether you win the recommendation. Visibility is the share of AI shopping answers where your product appears. Win rate is the share of those answers where it lands first. Position is where it ranks when it shows up. Mentioned price is what the engine quotes for your product, checked against your catalog price so you catch anything stale or wrong. Attributes are the dimensions ChatGPT uses to compare products in your category, ranked by how often each one comes up.

The full picture on every product

Click into any product and you see where AI sends buyers when it recommends you, whether that is your store, a marketplace, or a retailer. You see how ChatGPT compares you against competitors and where your product descriptions fall short of the dimensions it actually weighs. You see which products are gaining placements this week and which are quietly bleeding them. And you see the exact prompts driving your visibility, alongside the ones you are not showing up for yet. The prompts you lose are your clearest list of what to fix.

A brand-level overview

The Overview rolls every product up to the brand level and answers the first question most teams ask: in my category, which brands and products show up most in AI shopping answers, and where do I rank? You get visibility, position, share of voice, and win rate over time for you and your competitors, the top performers at a glance, and the shopping queries driving those answers, including the ones gaining traction and the new ones just appearing. Those queries overlap heavily with Google Shopping search, so they double as a feed for your wider search and merchandising work.

Connect your catalog in minutes

Three ways to bring products in: paste your Shopify store domain and Peec pulls the catalog from your public product feed, upload a Peec CSV for non-Shopify catalogs, or connect an existing Google Merchant Center feed. Your products go live in minutes, organized by category and filterable across the Shopping overview, and they are matched against the last 30 days of chats so your metrics start with history instead of from zero. The docs walk through each option.

Chats are the raw evidence of how AI engines answer on your topics. Finding the ones that matter used to mean scanning them one at a time, with no way to filter by what was retrieved or who got mentioned.

It is out of preview now. You can filter chats by brand, see mention and source counts as sortable columns (citations are off by default to keep the view clean), and remove chats with a delete toggle. Pending chats are back in the table, it loads noticeably faster on large workspaces, and chats from deleted or archived prompts no longer appear, so what you see reflects your live prompt set. Sort by mentions or sources to surface the chats worth opening first.

Three new ways to connect your logs to Crawl Insights

Crawl Insights tracks how AI crawlers like GPTBot and ClaudeBot move through your site. Without a dedicated integration, getting your log data in meant manual file uploads, an ongoing effort that was easy to let slip. You can now connect through a WordPress plugin, AWS CloudFront, and Google Cloud, on top of the Vercel, Akamai, and generic webhook options from earlier. Point your logs at Peec once and AI crawl monitoring runs on its own, wherever your site is hosted. The setup docs walk through each option.

Bulk tag editing for large workspaces

Structuring a big workspace used to mean creating and fixing tags one at a time, which got tedious after a large prompt upload. You can now create up to 50 tags at once, and select multiple tags in the table to recolor or delete them in one go.

Brand Insights graduates from Early Access to everyone, in a dedicated Brand category in the nav. It gives you a per-brand deep dive on visibility, sentiment, position, and share of voice, with full time series and a heat map, so you can see how any brand moves over time and where it stands right now.

Pick any two of brands, topics, tags, or models for the x and y axes and outliers surface fast. The strongest-model and weakest-model highlights are the quickest way to decide where to focus: you can see at a glance which engine is carrying your visibility and which one is leaving it on the table.

Movers on Domains and URLs

Both the Domains and URLs pages now open with a bar chart split into Top, Trending, Losing, and New tabs, so you can see which sources are gaining or losing ground without scrolling the table. Click any tab for an expanded view of up to 50 entries. It is the fastest way to catch a competitor breakout or a sudden drop the moment it happens.

Custom URL and domain classifications

You can now create your own URL types and domain types and assign them to sources, on top of our default classifications. The whole taxonomy is available end to end in the Customer API and through MCP, so you can script your own classification scheme or wire it straight into an agent workflow.

A cleaner way to read chats

We rebuilt the chat detail view that opens whenever you click into a chat. The layout is cleaner, the hierarchy between answer, sources, and features is clearer, and you can now move from one chat to the next without closing the view. Reading through chats one after another is something a lot of you do, and this makes it much faster. The Prompts table also gained chat feature columns: web search is shown by default, and you can switch on the rest from the column picker to filter down to, say, only the prompts where the model actually went to the web.

More of your data through MCP and the API

Chat features are now queryable on individual chats and available as a filter on list_chats: ads, maps, product carousels, web search, image galleries, and shopping fanouts. You can ask Claude to "show me every chat where ChatGPT served an ad for my brand in the last 30 days" or filter the same way in your own scripts, and get_chat returns the full detail including ad payloads and map markers. Crawl Insights is now exposed the same way, so agent and bot crawler data is queryable end to end. We also added Personal Access Tokens you can create right on the API Keys page for use with MCP, and retrievals delta plus an onlyNew filter on the public source endpoints so you can pull just what changed in a period.

Chinese models

We added DeepSeek and Qwen, so you can track visibility across the Chinese AI ecosystem alongside the engines you already follow.

Early Access

Activate Early Access in your company settings to turn these on yourself.

• Gap Analysis is now its own page in the nav instead of a toggle inside Sources. Same feature, much shorter path to it.

• Source Bookmarks let you bookmark URLs and domains straight from the sources tables and filter to show only the bookmarked ones, so you can keep a curated set of references in view across reports.

• Command/Ctrl-K opens a keyboard-driven search palette from anywhere in the app, with nested search into prompts, domains, and URLs. This is the first version, with more coming.

• Instant Pitch Projects let agencies spin up a pitch project that runs 10 chats per prompt instantly, instead of collecting them over seven days, so you can have data ready for a pitch the next day.