3RD x Cernel: Find the Products AI Skips

3RD and Cernel partner: find the products AI shopping skips, and fix them across your whole catalog.

3RD and Cernel have partnered so retailers and brands can see which of their products AI shopping answers leave out, add the product details those answers look for across the entire catalog, and measure whether the products start showing up.

3RD is the AI visibility platform. Cernel is the Danish AI platform for e-commerce product data. One finds the gap. The other fixes it at catalog scale. Here's how that works in practice, for the shopper, for your team, and for the products in between.

What the shopper sees

Someone asks ChatGPT for "a waterproof running jacket for women".

They don't get ten blue links. They get a short list of product cards: a name, a price, a rating and the shops that sell it. To build that list, the assistant runs its own searches behind the scenes, looking for the details that decide the answer: how waterproof, how breathable, what fit.

The shopper never sees your category page. They see the shortlist. If your jacket's product data doesn't carry those details, your jacket isn't on it. Even if the same page ranks well on Google.

Enlarge

What the shopper sees: a shortlist of product cards, and your jacket isn't on it.

Why this is hard for retailers today

Most e-commerce teams run their product data the way they always have: a feed built to Google Shopping's specs, supplier spreadsheets, and manual fixes when something breaks. That setup has three blind spots in AI shopping:

• You can't see it. Your analytics show traffic and rankings. They don't show which products an AI answer named, which it skipped, or which shop it sent the buyer to.
• You don't know what's missing. Even with the answer in front of you, it's not obvious which detail kept your product out.
• You can't fix it by hand. Adding and standardizing attributes across tens of thousands of products, dozens of suppliers and several languages isn't a plan. It's a year.

How the partnership works

The work runs as a loop: measure, diagnose, fix, measure again. 3RD handles the first two steps and the last one. Cernel handles the fix.

Step 1: See which products AI shopping answers skip (3RD)

3RD's Product tracking reads the shopping answers to the prompts you track and records every product card in them: which product was named, at what position, at what price and rating, and which shops carried it.

Upload your catalog as a CSV and 3RD matches it against everything the answers named, by GTIN first, then MPN, then attributes. That gives you three lists:

• your products AI recommends, and how often
• your products AI knows about but rarely recommends
• products AI keeps recommending that you have nothing like

Start with the second list. Filter to your products and sort by visibility, lowest first. A product AI already knows about is usually a cheaper fix than getting a new one recognized at all.

If you're a retailer rather than a maker, switch the view from Brands to Shops. Your question isn't whose product was named. It's whether the answer's link pointed at you or at a rival stocking the same product. Look for your own products where the Retailer competition column is full of somebody else. Those are winnable without changing anything about the product itself.

Enlarge

3RD Product tracking, filtered to your own products and sorted by visibility, lowest first.

Step 2: See what the answers are looking for (3RD)

Every product in 3RD has its own record: the shops that carry it and at what price, how it competes across shops, and the shopping queries that surface it most. The response inspector opens the captured answer itself, with the product blocks beside the text.

Put a product that wins a topic next to yours. Often the gap is a detail you simply don't carry: a waterproof rating, a fit, a material. Sometimes it's the words. Your supplier calls the color "midnight blue", and the shopper asks for navy. The details and terms the winning product carries that yours doesn't are your enrichment brief.

Enlarge

The response inspector: the ChatGPT answer as it was captured, with the product card beside it.

Enlarge

Competing products: what else the same answers named, which shops carried it, and how visible it was.

Step 3: Fix the data across the whole catalog (Cernel)

This is where Cernel takes over. You set the rules once, and Cernel applies them to every product they fit:

• Every product gets a category. Cernel classifies the whole catalog into your taxonomy, or a marketplace's, and each category decides which attributes its products need. Add "weather resistance" to outerwear once, and every jacket in the category gets it.
• Search demand points to what's missing. Cernel looks at what shoppers search for in each category and recommends new attributes, or new values for the ones you have. It's a useful second opinion next to the brief from 3RD.

Enlarge

Cernel's Market insights: each category measured against what buyers search for, with completeness and open recommendations per category.

• AI agents fill the gaps. Each agent owns one attribute: a description, a material, a fit, a color mapped to the term shoppers use. Agents follow your brand guidelines and product rules, and can only pick from the values you allow, so the output stays on brand and on spec. External data agents research the web for details your supplier never sent, such as specs or GTINs, and cite the sources they used.
• You decide what needs a review. New values can go straight live, or land in a review queue first, next to the current value with the agent's reasoning and the data it used. From there, your team approves, edits or rejects them.
• It writes in every language you sell in. Each market gets copy written in its own language, with local headings and terms, rather than a word-for-word translation.
• It pushes back to where your data lives. Shopify (including metafields), Struct and Inriver, product feeds such as Google Shopping, CSV, or your own systems over the API. Automations then classify and enrich new products as they arrive, so the fix doesn't turn back into a backlog.

Enlarge

Cernel sorts every product into a category, and the category decides which attributes it needs. Here, outerwear gets weather resistance, insulation, warmth and hood type.

One brief from 3RD becomes a catalog-wide change in Cernel, with tens of thousands of products in a single run, instead of a spreadsheet someone works through by hand.

Enlarge

One run across the whole catalog: pick the attributes and languages, and decide whether values go to review first.

There's a bonus for Step 1. 3RD matches your catalog by GTIN first, so every missing GTIN Cernel finds is one more product 3RD can recognize exactly in AI answers.

"Most product data was written for a filter menu and a search box. AI shopping assistants read it differently: they look for the exact detail the shopper asked about, and if it isn't there, the product doesn't make the shortlist. 3RD shows which details are missing. Cernel adds them across the whole catalog, in every language, and keeps adding them as new products arrive."
Magnus Bruun, CCO and co-founder, Cernel

Step 4: Measure whether the products start showing up (3RD)

Cernel records when each value was last changed and who changed it. That date is your "before and after" line. Re-upload the enriched catalog to 3RD, which re-runs the matching, then read product visibility per topic, with its change against the previous period. A product's visibility is its share of the captured answers that named any product at all, so products are directly comparable.

Three things keep the result honest:

• Read the trend, not the day. AI models are probabilistic, so the same prompt can return different answers on different days.
• Give it time. Models re-crawl on their own schedules, so expect days to weeks between a change going live and a change in the answers.
• Look beyond your own shop. AI shopping answers lean heavily on merchant listings. If you're a brand, the shops that carry your products need the enriched data too, so send it to them.

Enlarge

Product visibility per topic in 3RD, with the change against the previous period.

Who does what inside your company

The split that works is marketing owns 3RD and the product team owns Cernel, with a clear handover between the two.

1. Marketing tracks the shopping prompts that matter, finds the products that are skipped, and writes the brief: which products, which details, which markets.
2. The product team turns the brief into attributes and agents in Cernel, and pushes the output live, with a review step wherever they want one.
3. Marketing reads the trend per topic from the date of the change, and reports it back to both teams.

Set up tracking in 3RD before you start enriching in Cernel. Without a baseline, you'll know the data got better, but you won't be able to show whether AI shopping answers noticed.

Who it's for

Retailers and brands with large catalogs, many suppliers or many markets. It fits fashion, beauty, health and home best. Makers read the Brands view, and shops read the Shops view.

What to know before you start

• Product tracking is in closed beta. It's switched on per organization. Ask us to enable it.
• A dash is not a zero. If no answer in your scope named a product, 3RD shows a dash, meaning unmeasured. Widen the date range or topic before reading it as absence.
• The result we measure is appearances. We'll show you whether the products you enriched start appearing in more answers. We won't claim the sales that follow. That depends on your price, stock and pages too.
• Your team publishes. Enriched data only counts once it's live in your store, your feeds and your product pages.

FAQ

Do I need both 3RD and Cernel?

No. Each works on its own. 3RD alone shows you which products AI skips. Cernel alone improves your product data. Together, you know which products to fix first, and whether fixing them worked.

Which AI assistants does this cover?

Product-level tracking reads ChatGPT's shopping answers today. For brand-level visibility, 3RD tracks the assistants you choose, including ChatGPT, Gemini, Perplexity and Google's AI Overviews.

Does everything have to be approved before it goes live?

No, it's your choice. Values can go straight live, or wait in a review queue with the reasoning and sources behind them until someone on your team approves them.

How fast will I see a baseline?

Your first 3RD run gives you a snapshot within the hour, and the full picture builds over 24-48 hours. Changes after enrichment take days to weeks to show up in answers.

Do I have to upload my catalog?

Everything in Product tracking works without it except catalog gaps. The catalog is what turns "these products get recommended" into "these products get recommended, and we have nothing like them".

Which languages and markets?

Cernel writes product data in every language you sell in. In 3RD, a market is a country plus a language, so Denmark in Danish and Denmark in English are tracked separately.

See which of your products AI skips

Book a demo. We'll show you which of your products ChatGPT's shopping answers leave out, and how Cernel fixes them across your catalog.

• Book a demo with 3RD →
• Book a call with Cernel →

Already tracking in 3RD? See how Cernel fixes the data at cernel.com.

Magnus Windeløv

Head of Marketing