Your Product Titles Are Written for Humans. The AI Reads Them Differently.

For years, product title optimisation has meant one thing: front-load the keywords humans search for. Brand, product type, key attribute, size, colour. Pack it in, keep it under 150 characters, move on.

That formula worked when the only thing reading your titles was a matching algorithm. But that's not the world we're in any more.

Google's AI Business Agent doesn't match keywords. It interprets meaning.

When a shopper asks "What's a good waterproof jacket for hiking in Scotland?", the agent isn't scanning your title for exact-match keywords. It's parsing your title, your description, and your attributes to understand whether your product is a credible answer to that question.

That's a fundamentally different job. And most product titles aren't built for it.

What the old rules optimised for

The traditional title formula was designed for a keyword-matching engine. It prioritised:

• Exact-match search terms in the first 70 characters
• Brand name prominence for branded queries
• Attribute stacking - size, colour, material - for long-tail coverage
• Character-limit efficiency over readability

This produced titles like: "Nike Air Max 90 Men's Running Shoes Black/White Size 10 UK Trainers Sneakers"

Functional. Keyword-dense. And completely useless for a conversational AI trying to recommend a product to someone who asked "What's a comfortable everyday trainer?"

What the AI actually needs

An AI agent interprets titles as semantic signals, not keyword lists. It's looking for:

• Product identity - What is this thing, precisely? Not just "jacket" but "insulated waterproof hiking jacket"
• Use case clarity - Who is it for and when would they use it? The agent needs to match products to intent, not just terms
• Differentiating features - What makes this product different from the other 47 jackets in your catalogue?
• Natural language structure - Titles that read like coherent phrases, not comma-separated keyword dumps

The agent doesn't care that you mentioned "jacket" three times. It cares whether it can confidently recommend your product when someone describes what they need in plain English.

The title isn't working alone any more

In the old model, the title did most of the heavy lifting. The description was an afterthought. Attributes were filled in if someone had time.

In an agentic model, the AI reads everything together. Your title, description, attributes, and custom labels form a single data profile that the agent uses to understand your product. A great title with thin attributes still produces a weak recommendation.

The title is now the headline of a document. Not the entire document.

That means your descriptions need to carry weight. Your attributes need to be complete. Your feed attributes need to answer the questions that shoppers will ask - because the agent will try to answer them using whatever data you've provided.

What "optimised" means now

Title optimisation in the agentic era means writing titles that help an AI understand your product well enough to recommend it confidently. That's a different brief.

• Lead with clarity, not keywords. "Lightweight Insulated Hiking Jacket - Waterproof, Breathable, 3-Season" tells the AI more than "Men's Jacket Waterproof Hiking Outdoor Black"
• Include use case signals. Words like "everyday", "travel", "performance", "office" help the agent match products to intent-driven queries
• Avoid redundancy. If the colour is in your colour attribute, you don't need it in the title. The AI reads both
• Write for recommendation, not just discovery. The agent isn't just finding your product. It's deciding whether to recommend it. Make that decision easy

The commercial implication

Brands that rewrite their titles for semantic clarity will get recommended more often in conversational shopping experiences. That's not speculation - it's how language models work. They recommend things they understand.

Brands that leave their keyword-stuffed titles untouched will still show up in traditional Shopping results. But they'll be invisible in the conversational layer - the layer Google is investing billions to build.

This is not a future problem. The AI Business Agent is already live. The UK rollout is a matter of when, not if.

Your product titles were written for a search engine. They need to be rewritten for a sales agent.

The brands that do this first will compound. The ones that wait will wonder why their Shopping performance is quietly declining.