What the Gemini API / Google AI Studio API actually does on this website

A few weeks ago we told you we rebuilt this site with AI sitting next to us the whole way. That post was the headline. This one is the detail. Specifically, it's about the features running on Google's Gemini API, what each one does, and what it actually changes for you when you land on a page.

No vague AI talk. These are real features, live on the site you're reading this on, and you can go and use every one of them the moment you finish here.

AskClicky. An assistant that has read everything.

Tap the chat bubble in the corner, or open the Ask Clicky tab inside Get Started, and you're talking to a Gemini model we've handed the whole agency to. Every service, every sector we work in, the case studies behind our results, who to speak to about what. It runs on Gemini's fast Flash model so replies come back in a beat, and it remembers your conversation as you move from page to page.

What it changes: the front door to the agency is open 24 hours a day. Quick question, quick answer. Bigger question, and it points you at the right person and the right case study rather than a generic contact form. No waiting until Monday.

Describe your project. Get matched in seconds.

Open Get Started and there's a second Gemini tool waiting. Tell it about your project in your own words and it reads the brief, then hands back the one to three services that actually fit, two or three case studies that look like your situation, and a sensible next step. We've fenced it in deliberately: it can only point at real Clicky services and real case studies, so it can't wander off and invent something we don't do.

What it changes: the bit of a project that used to need a 30 minute scoping call now happens the second you've finished typing. You get a route through the agency that's relevant to you, not a tour of everything we offer.

Talk, don't type.

Briefs are easier said than written. So next to the box you can hold a button and just talk, up to five minutes of it. Gemini transcribes the recording into clean English, and there's a polish step that tidies the grammar, drops the ums and the false starts, and groups your points together while keeping it in your voice. Then it runs the same matcher over the result.

What it changes: the friction of starting a conversation drops to near zero. It's faster for anyone in a hurry, kinder to anyone who finds a blank text box intimidating, and it tends to produce a far richer brief than the three lines people usually type. Better in for you, better understanding for us.

Every page has artwork that didn't exist a month ago.

The visuals are Gemini's family too. Gemini writes the creative brief for an image, its image model renders the service heroes and icons you see across the site, and Google's Veo video model takes a still and animates it into the looping hero films: the balloon on our anniversary post, the moving moments elsewhere. Each service page now gets its own piece of art instead of sharing a stock category image with five others.

What it changes: a distinct visual identity on pages that used to make do with whatever was nearest. And motion that would once have meant a shoot day and a budget now ships in an afternoon, which has quietly raised our own bar for what a good hero looks like.

Press play on anything.

Every blog post has an audio read-through, and our case studies and monthly recaps come as short two-host podcasts between Laura and Ollie. Gemini drafts the script for those (a natural British back-and-forth that pulls out the metrics and the story) and ElevenLabs gives the hosts their voices. It regenerates automatically whenever new work goes up.

What it changes: our writing stops being a screen-only thing. A case study can be listened to on a commute, a month of news can be caught up with on a walk. No app, no paywall, just a play button.

Why Gemini, and the honest part.

We chose Gemini's Flash models for the live features because they're quick and cheap enough to run for every visitor without thinking about it, and because we can make them answer in a strict, structured format. That last bit matters more than it sounds. When the project matcher has to return its answer in a fixed shape, drawn only from a list of things we actually do, it stays on the rails. It recommends, it doesn't hallucinate.

And the honest part, same as last time. Gemini didn't decide our strategy, write our brand voice, or choose which case studies are worth showing you. It does the legwork. It reads the brief, it drafts the script, it renders the picture, it answers at two in the morning. The judgment underneath all of it is still ours.

And it's already going into client work.

None of this is a one-off for our own site. An assistant trained on a client's own services and FAQs, sitting in the corner of their site. A matcher that routes their visitors to the right page. Audio versions of their content. AI-generated hero art for brands that want premium motion without the production cost. It's the same toolkit, and our teams are shipping it into client projects right now.

If any of it looks like a fit for you, tell us about your project, or just ask AskClicky in the corner. Fittingly, it can probably explain the rest.