I stress-tested Atlas on two things that normally eat time:

1) Shipping something technical I’d been avoiding.

I opened a clean repo and asked Atlas to scaffold a tiny backend: minimal REST API, auth, Postgres, Dockerfile. It generated the file tree, starter models and routes, and the bash commands to run locally. When I hit a 500, I pasted the trace into the sidebar; Atlas spotted a type mismatch, rewrote the model, and re-ordered the migration. No whiplash between “docs,” “YouTube,” and “my code.” It stayed with me across GitHub, package docs, and local logs, like a patient senior who doesn’t mind naïve questions. I still used judgment; the friction just melted.

2) Going full agentic shopper.

I gave it a real brief: black-tie capsule, strict budget, sizes, delivery to Sydney by a hard date, no polyester, backups if a size sells out. It drafted two style routes, cross-checked shipping cut-offs to my postcode, and suggested alternates with faster couriers. I pasted measurements and a buried size chart; it translated them into likely fit, hem estimates, and a heel height that keeps the line right (and cobblestones in mind). The research heavy lift? Off my plate.

That’s the quiet revolution: discovery, comparison, and action collapse into one conversation. Websites become data sources, not destinations. And if your data is messy or missing, you disappear at the moment it counts.

Not fear - new literacy

Simon Beard wrote about this “ChatGPT browser moment” and why he’s switching: change the default, change the market. If the default browser is AI-first, answers happen inline and execution sits inside the result. Demand capture shifts from classic SEO to what we call AIO - AI Engine Optimisation. I agree. But my takeaway isn’t doom. It’s a nudge: learn how assistants read.

Think of Atlas as a very motivated, very literal intern. It will assemble my journey from whatever truth it can find. Give it structured truth and it becomes your best salesperson. Starve it or contradict yourself and it quietly routes around you.

So… what replaces SEO?

SEO doesn’t vanish; it evolves. The new stack is AIO + GEO:

AIO (AI Optimisation): Make your brand legible to assistants. Use structured data for products/services (materials, variants, pricing, availability), consistent naming across your site and the high-authority places models trust, and short “why it works” blurbs the AI can quote without guessing.

GEO (place/local optimisation): Make your locations legible. Store-level pages with clean LocalBusiness schema, accurate hours, phone, and geo coordinates; postcode-aware delivery windows; even a “frequently updated” inventory signal. Assistants care about place the way people do: Can I get it here, now, reliably?

When I asked Atlas to dress me for an event on a deadline, it didn’t just show dresses. It reasoned about fabric + weather + fit + delivery SLAs + my route to a nearby store. That’s where AIO and GEO meet: structured truth + place truth = visibility inside the conversation.

A tiny checklist I wish every brand shipped yesterday

Keep it boring. Boring is what assistants love.

Variant-level truth - expose size/color as fields with stock status per variant.

Measurements, not vibes - garment dimensions alongside size labels, consistent site-wide.

Shipping as data - methods, postcode cut-offs, ETA bands as fields (not buried paragraphs).

Returns in one line - window, method, restocking rules; something quotable and certain.

Store pages that are actually useful - address, hours, phone, geo coordinates, accessibility notes; tie inventory signals to locations if you can.

Capsule pages - “Black-Tie Capsule,” “48-Hour City Kit,” “Room Type Matrix,” with reasoning blurbs the model can lift.

Kill contradictions - sync names, specs, and prices across your site, marketplaces, and review profiles. If the AI has to reconcile diffs, you lose time and placement.

Do these and you don’t have to shout. The assistant does the quiet work of putting you in my shortlist because you’re easy to reason with.

Where Berland sits in this shift

Our whole thesis lives here. Berland is the hybrid service-platform that:

Audits what AI already “knows” about you across your site and the third-party sources it trusts.

Fixes the diffs (the contradictions that make assistants hesitate).

Structures your truth (AIO) and makes your places legible (GEO).

Monitors how often and how well you appear inside AI answers over time.

We’re not replacing your brand voice or creative. We’re making sure the assistants can find, understand, and recommend you - automatically.

Closing thought

Atlas didn’t make me magically senior. It just made the web cooperative again. I can be scrappy without being sloppy. That’s not something to fear; it’s something to learn - and it’s learnable fast.

If you want to be the obvious answer inside this new browser reality, that’s the work we do at Berland. Let’s make the assistant choose you.