Pomelli is live in Switzerland, and it ships a free audit of your site
Pomelli is now available in Switzerland, and in more than 195 countries overall. The read is the same everywhere: yet another free campaign generator for small businesses, this one signed Google. There is something more interesting inside. To build those campaigns, Pomelli starts by reading your website and extracting a brand identity from it, which it then shows you. In other words, Google is handing out, for free, the report of what a machine understands about your site. That is the product.
Pomelli in one minute
It is a Google Labs tool, built with DeepMind. You give it your website address, it analyses the site and derives what it calls a “Business DNA”: your colours, your typefaces, your values, your tone of voice, your imagery, your product catalogue. The analysis takes up to eight minutes.
Then you describe an intent in one sentence, something like “promote my new collection”, and it returns several campaign concepts with matching visuals and copy, ready to be adapted for Instagram, Facebook, LinkedIn, TikTok and YouTube. It is free, it needs a Google account, and it is restricted to adults.
For positioning: it goes up against Canva on social creative, with a brand engine in front of it. Search is outside its scope.
Available almost everywhere, in English only
Pomelli launched in late 2025 as a test in four markets only: United States, Canada, Australia, New Zealand. In March 2026, Google opened it to more than 170 countries. Today the official help page listing territories names over 195 of them, Switzerland included.
Now the part the announcement coverage leaves out: the output is in English, and English only. That applies to the generated content, visuals and copy alike. No date has been communicated for other languages. Google also notes that certain features and image types remain unavailable depending on the country, with image generation the most constrained part in the European Union because of the AI Act and the GDPR.
For a Swiss small business, that means English visuals and hooks aimed at customers who speak French, German or Italian. Reworking all of it amounts to redoing half the job, and precisely the half that requires judgement. A company serving three language regions multiplies the problem by three. There is a certain irony in a monolingual tool serving a four-language country worst of all.
The Business DNA is an audit dressed up as a feature
Here is where it becomes useful. To generate anything at all, Pomelli first has to understand. It crawls your site and derives a reading from it: what you are for, who you speak to, in what tone, with what visual identity. That reading is shown to you, and you can correct it.
The mechanism echoes ED’s brand voice: a machine reads what you publish, infers who you are, then uses that to write in your place. Pomelli has the good sense to show the intermediate step.
The value is in the reading. You get, free, in eight minutes, what a Google machine understands about your site with nobody helping it along. Correcting the profile by hand inside the tool costs you exactly that benefit. And if what it shows comes out wrong, hollow or generic, the site itself is what needs looking at.
The case to watch for: a very precise expertise that comes back as “professional services”. When that happens, the home page is what is at fault, staying too vague about what is actually sold.
Why it matters more than a post generator
Because it is exactly the same mechanic that decides whether a model cites you. Read a site, extract an entity, a field of competence, a positioning: that is what Pomelli does to compose a campaign, and it is what generative engines do to pick their sources when somebody asks a question in ChatGPT, Gemini or Perplexity. I covered that mechanic in detail in the article on GEO.
The correspondences are fairly direct. If the Business DNA comes back generic, your positioning is unreadable to automated parsing. If the values come back hollow, your site is a stack of interchangeable adjectives. If colours and typefaces shift from one page to the next, you have a design consistency problem that humans notice too. If the catalogue comes back incomplete, your product pages are badly structured.
Those conclusions are well known. What is new is getting them free, in a few minutes, in readable form, on the infrastructure of the very party that ranks you.
How I would use it
Run it on your site and read the Business DNA exactly as it comes out. Note the three biggest gaps between what the machine understood and what you actually sell. Then fix the site rather than the profile: the profile regenerates in eight minutes, whereas your home page also works for classic search, for the models, and for people. Run it again afterwards and compare. If the gap narrowed, the fix landed.
For the generation itself, treat the visuals as drafts and rewrite the copy in your own language. It is still faster than a blank page, and that is about all it is worth today. For a publishing rhythm that holds up over months, in your own language and pushed to the channels on its own, ED · Editorial is the tool for the job.
One caveat, though. Giving up your site URL only shows what Googlebot already sees. Everything you add afterwards inside the tool, catalogue, briefs, positioning material, goes to an experimental Google Labs product. Stick to what is already public, which counts double if you handle client data covered by the Swiss FADP or the GDPR.
What it says about the moment
Google has just put a tool in the hands of any small business that shows, in eight minutes, how a machine reads its website. That is a side effect: the stated intent is producing campaigns. It remains the most useful piece of information the tool delivers.
You can look at Pomelli as one more post generator, and that would be accurate. You can also read it as the signal that machine readability has just become a visible property of your website, in the same way its loading speed is. That is also the principle behind SEO Cartograph: look at your own site the way a machine looks at it, before deciding what to fix.
The difference between those two readings is roughly the one between the people who will run the test and the people who will read a summary of the tool.
Also worth reading: Generative Engine Optimization (GEO): how AI picks the sources it cites · AI adoption in business: why model power is not enough
Thomas Nedjar has spent fourteen years in digital acquisition and e-commerce. An entrepreneur, he founded and ran several e-commerce companies between France and Switzerland, with solid experience in international and intra-EU trade. He is now Senior SEO/GEO, AI & Tech Expert at Suisseo, and builds his own marketing applications: ED (automated community management) and SEO Cartograph (local SEO audit crawler). A speaker (CVCI) and trainer, he shares his hands-on insights on SEO, GEO, automation and e-commerce here.
All articles · LinkedIn