The questions product data teams ask us.

Cernel is an AI platform for product data. You bring in products from your shop, PIM, supplier files or a feed, and Cernel places each one in your categories. AI agents then fill in what is missing: descriptions, SEO fields, structured attributes, translations and images, written by your rules.

You choose which values go live automatically and which wait for a person to approve. Approved values sync back to the shop and PIM systems we integrate with, or leave as a CSV export, a Google Shopping feed file or over the API. See the integrations for where data comes in and goes out, or click through it yourself.

Cernel works with your PIM as the AI layer that fills it: it takes products in, enriches them and sends approved values back to the fields you mapped. The PIM systems we integrate with connect in both directions; any other PIM, your ERP or another source of truth connects through the API or files.

If you have no PIM, Cernel works with your shop or with CSV files.

E-commerce, master data and PIM teams at brands and retailers with large or growing catalogues, in fashion, beauty, sports, home, B2B and industrial parts. The value grows with the number of new products, categories, languages and channels you handle.

In a demo we look at your catalogue and volumes together, so you can see what Cernel would do for your team. See who runs on Cernel.

Yes. An automation enriches the products in a category as they arrive, or again when the fields you choose change, with the agents you picked. Fields you set to go live automatically sync to your connected shop or PIM straight away; the rest wait for approval in Review.

Yes. Most teams run their back catalogue once, category by category, and then let automations handle new products. With "only enrich missing fields" on, Cernel fills what is empty and leaves the values your team already wrote alone.

A single enrichment job takes tens of thousands of products and runs in the background, so your team can keep working. Most products are done in under a minute; images and web research take a few minutes per product. Each plan includes a monthly enrichment volume that you can follow on the usage page.

Yes. Several shops can connect to one account, and a product with the same SKU or barcode in two stores becomes one product in Cernel, so its facts are filled once. Texts that should differ per store, such as a description in each brand's voice, get their own fields and agents.

Brands you want to keep fully apart can run as separate organisations, with one login and a switcher between them.

Yes. New accounts start with every value waiting for approval in Review, where you approve, edit or reject each one, or many at once. Once you trust a field, you can set it to approve automatically, while fields such as safety or compliance information keep a person in the loop.

You choose which values go live automatically and which get a review step; review steps check every value and only ask a person when something fails or needs a decision. More on review steps.

Agents work from your product data, your supplier files, the reference material you give them and, when you turn it on, the websites you allow. For a field with a value list, you choose whether the agent must always pick a value or may leave it empty when the data does not say.

Every AI-written value shows its inputs, the agent's reasoning and the prompt, so you can see where it came from. You test an agent on your own products before it touches the catalogue, and a review step can hold back any value that breaks the rules you write in plain words.

From what you send, such as supplier files, PIM or ERP fields and product images, and, when you turn it on, from the web. Research agents look the product up by EAN, GTIN or name, only on the websites you allow, and show their reasoning with every value so a reviewer can check it.

For regulated claims, such as in beauty or health, we recommend making your own approved texts the source.

Yes. Agents can use the product's images as a source, so anything a person could see in the photo, such as colour, pattern, shape or neckline, can fill an attribute. Facts a photo cannot show, such as material composition, come from your source data or, with web research on, the websites you allow.

Yes. Agents read your tone of voice, brand rules and terminology, including words to avoid, from shared reference material, so the same guideline applies to every product. Update it once and every agent that uses it follows. Words that must never be translated go into each language's own instructions.

No prompt engineering is needed. Each attribute has its own agent, and you tell the agent chat in plain language what it should write or choose. Cernel writes the agent's instructions from that, and you test and adjust it on real products before it runs on the catalogue. Every change is saved as a version you can go back to.

Agents can build on each other, for example a short description written from the long one, and a field set high in your category tree applies to every category under it. See product data enrichment.

Yes. Cernel places every product in a maintained category tree from day one, and you can add your own, such as your shop's or a marketplace's, and let Cernel classify into those too. Attributes can be free text or chosen from your own value lists, are set per category and are inherited by the categories below.

See categories and fields.

Yes. For a field with a value list, the agent only chooses from the values you allow, so every supplier's "navy", "marine" or "dark blue" ends up as the one colour your filter shows.

Only where you tell it to. Cernel sends only the values it enriched, or that your team edited in Cernel, to the fields you mapped; data it only imported stays as it is in your systems. Every value keeps its history in Cernel, so you can see and restore an earlier one, and you can export your catalogue as CSV at any time.

Cernel connects in both directions to the shop and PIM systems we integrate with. On top of that there is CSV import and export, a Google Shopping feed you can import and export, and an API for every other system, such as an ERP, another PIM or your own tools.

Approved values go back to the connected shop and PIM automatically, typically within a minute. Every other system collects them over the API or takes an export. See every integration.

A product feed or a CSV export with names, EANs and whatever attributes you have is enough to start, and an integration or the API takes over for ongoing work. Several sources can be matched on SKU or EAN, and an EAN helps the agents find the right product on the web.

Add your tone of voice and brand guidelines, and the agents have what they need.

Yes. Any other system connects through the API: you send products in, and your platform collects the enriched values. See the API docs.

Yes, over the API. Your ERP, or a small piece of code next to it, asks the API for the values that changed and writes them in. A CSV export works too.

You connect your store with Cernel's app, and Cernel picks up new and changed products, typically within a minute. Approved values are written to the fields you map: title, description, SEO title and description and your own custom fields, which Cernel can create for you, plus the images its agents generate. Each language syncs as a translation in your shop.

Yes. Each channel or store can have its own fields, such as marketplace bullet points, a feed title or a description for a B2B shop, each with its own agent and rules. They are written from the same product data, so the facts stay the same everywhere.

Yes. Cernel gets the data ready for each channel. You add the channel's category tree, agents fill the fields it requires, such as feed titles, and you export it as a Google Shopping feed file, a CSV in the channel's own columns or over the API, into the feed or listing tool you already use.

See marketplaces.

CSV files in whatever layout and column order they come, with images as links, and a Google Shopping feed by link or as a file. Before anything is written, Cernel shows how many products the file will create and update.

Cernel reads each supplier's file as it comes and suggests where every column belongs in your catalogue, so your team checks it instead of retyping it. Next time, the last import from that supplier is the template, and only new columns need a look. Files about the same product join into one product by SKU, EAN or your own product number.

See supplier data onboarding.

Yes, both the main product and each variant. Variants come in with their main product from your shop, your PIM or the API, and files can update their values. Enrichment includes variants by default, and an agent can use the variants' values, such as every colour on offer, when it writes for the main product.

Yes. Cernel exports a Google Shopping feed file, ready to upload to Merchant Center, and you export a fresh one whenever your data changes. Values outside Google's fields go into the custom labels. See product feeds.

Google and AI assistants decide for themselves what they show. What you control is your product data, and they match products on its attributes, so every attribute that is filled in, such as material, fit, size or use, is one more question your product can answer. Complete, structured product data is exactly what Cernel delivers, with meta titles and descriptions written for what shoppers want to know, not stuffed with keywords.

Bring your own keyword lists as a source, or let search demand show, from monthly Google search figures, which attributes shoppers search for that your data is missing.

Google ranks content on how helpful it is, not on whether AI wrote it. Cernel writes each text from that product's own data and your rules, so pages do not repeat each other, and your team can approve what goes live. You can roll out one category at a time and compare before and after.

Every language you sell in, including regional variants such as Brazilian Portuguese or British English, and each run can write all of them. For each language you choose to write it from scratch, translate it from another language, or skip it, with its own instructions for grammar, tone and terms that must not be translated.

A translation reuses the rules of the text it comes from, such as length, tone and format, and goes through the same checks. See translations.

Yes. Media agents create images, such as lifestyle and styled product shots, from the product's data, its images and your brief, in the format, size and file name you set, and write the alt text for the images they create. You can review each image beside its product before it syncs.

Research agents can also find a product's photos on the websites you allow, for you to review. See images.

Yes, Cernel is GDPR compliant. Cernel ApS is a Danish company, so your data is protected by EU data protection law: we process personal data in line with the GDPR and the Danish Data Protection Act, under the supervision of the Danish Data Protection Agency.

Every sub-processor works for us under a data processing agreement, and data that leaves the EU is covered by safeguards such as the EU's standard contractual clauses. Read how we handle data in our privacy policy.

The Cernel platform runs on Google Cloud in Frankfurt, in the EU, and that is where your product data and images are stored, protected by access controls, encryption and secure backups.

The services that work on your data for us, such as the AI models that write the values and the web search behind research, are on our sub-processor list, each with its purpose and location. The shop app works with product data only; customer data, orders and payments stay in your shop.

Yes. Write to mb@cernel.ai for the data processing agreement and the security documentation your IT or legal team needs, and we are happy to join a call with them.

For text and select fields you choose Google, OpenAI or Anthropic for each agent, and Cernel picks the model. Images are generated with Google's or OpenAI's image models.

The cloud services that run these models are on the sub-processor list, with where they process data.

Each organisation has two roles: admins manage users, integrations, API keys and platform settings, and users work with products, agents, enrichment and review. Invite as many colleagues as you need, by email. API access runs through service accounts, one per integration, with keys you can rotate without downtime.

A fixed monthly fee, set by the scope you need: catalogue size, how many new products you add, how many attributes and languages, and the features you use, such as review steps. The plan includes a monthly enrichment volume that fits that scope, so you are not charged per run.

We size it with you after a short look at your setup. Book a demo to get a price for your catalogue.

You start with a demo. You can click through an example right away, and in a demo we run it on products from your own catalogue. Your account is then set up with you during onboarding.

Typically 2 to 8 weeks from signing to fully set up, depending on how many categories, languages and integrations you start with. A dedicated onboarding specialist sets the platform up with your team, starting with the categories and attributes that matter most.

You can start on a CSV or feed import while the integration is built, so the first enriched products come early. See how teams got there.

Less than most teams expect. Onboarding is included, and the platform helps build the agents, but it needs your team's knowledge of your products and your standards. One person on your side, part of their week, is usually enough during set-up. After that, your team mostly looks at the values that wait in Review.

During onboarding you work with a dedicated onboarding specialist. After go-live, our customer success team supports you through in-app chat and regular check-ins, and passes your requests to the product team. docs.cernel.com covers every feature and integration.

You can, and for very low volumes a do-it-yourself setup can work. What we see is that in-house builds take months and usually cover one step, typically writing text, while the hard parts stay manual: reading supplier files, classifying into your categories, keeping quality consistent across thousands of products, translations, images, review and syncing back. Then someone has to keep it running as models change.

We wrote a fair comparison of both paths: Build versus buy for product data enrichment.

ChatGPT is good at writing one text at a time. Cernel is built for a whole catalogue: it connects to your shop, PIM or supplier files, knows your categories and attributes, applies the same rules to every product, shows the reasoning behind each AI-written value, lets your team approve values before they sync and sends them back to your systems.

You set an agent up once per field and run it across thousands of products and every language.

Built-in AI features usually write one text at a time, started by hand on a product that is already in the system. Cernel works before and around the PIM: it reads supplier files, classifies products in bulk, fills structured attributes as well as copy, translates, creates images and sends approved values back, by itself to the PIM systems we integrate with and through CSV or the API to others.