AI Brief: Steering Google's Ad AI With Your Own Words (2026)

Adil Ziani

Ecommerce Google Ads Consultant, ex-Google, Google Partner

Why automation without steering wastes budget

Here is the pattern I see in almost every account audit. An advertiser enables AI Max, leaves the steering blank, and lets Google's AI guess what the business is about from the website and the feed. The AI guesses reasonably well, maybe 80% right. The problem is that the remaining 20% is where the budget leaks: ads promising discounts that do not exist, matches on searches for products you do not stock, traffic from bargain hunters when you sell premium.

AI Brief exists to close that gap. It is a natural language interface, powered by Gemini, where you give AI Max the context it cannot scrape: your positioning, your constraints, your ideal customer. On the accounts I manage, writing a proper brief is now the first task after enabling AI Max, before touching bids or budgets.

The three guideline types, explained

Messaging Guidelines control what your ads say. This covers both directions: claims the AI should lean on, and claims it must never make. If you never discount, say so. If your warranty is your differentiator, say so.

Matching Guidelines control which searches your ads align with. AI Max matches beyond your keyword list, so this is where you draw the boundaries: categories you serve, categories you do not, intent levels you want.

Audience Guidelines describe who you want to reach: the customer profile, the use case, the price sensitivity. This shapes both matching and ad copy tone.

On top of these, text disclaimers guarantee that required text always appears in your ads, even when Final URL expansion sends users to pages you did not pick. For regulated categories or legally required mentions, this is not optional. Set it on day one.

Well-written vs vague: a real ecommerce example

Take a store selling organic skincare. A vague messaging guideline looks like this: "We are a premium natural skincare brand focused on quality." Gemini can do almost nothing with that. Every skincare brand on earth could write the same sentence.

A useful version: "All products are COSMOS-certified organic. Never claim medical benefits or that products treat skin conditions. Never mention discounts or sales. Always highlight free shipping over 50 euros and the 60-day return policy. Tone: expert but warm, no exclamation marks."

Same logic for matching. Vague: "Match searches related to skincare." Useful: "Match searches for organic and natural face care, serums and moisturizers. Do not match searches for makeup, hair products, men's grooming or medical skin treatments. Prioritize purchase-intent searches over informational ones."

The difference is constraints. The AI already knows you sell skincare. What it needs from you is everything it cannot infer: certifications, exclusions, policies, tone.

A template structure you can copy

This is the skeleton I use when writing briefs for client accounts. Adapt each block to your store:

• Who we are: one sentence, category plus differentiator. Not a mission statement.
• Must say: 3 to 5 proof points with numbers. Shipping thresholds, return windows, certifications, guarantees.
• Must never say: forbidden claims, forbidden words, competitor names, promotions you do not run.
• Match these searches: product categories and intent types you want, described in plain language.
• Do not match: adjacent categories you do not serve, free-seekers, job seekers, DIY queries.
• Our customer: two sentences on who buys and why, including price positioning.
• Disclaimers: any legally or brand-required text, added as text disclaimers, not buried in messaging notes.

Keep the whole brief tight. A focused page beats a rambling document. Gemini follows clear constraints far better than long prose.

Iterate with the search terms report

A brief is not a set-and-forget artifact. Every week, I open the search terms report on AI Max campaigns and ask one question: which of these matches would a good account manager have refused? Each bad cluster becomes a new line in the matching guidelines, plus a negative keyword as a hard backstop. Each ad rendering that misses the point becomes a sharper messaging line.

After three or four cycles, the matches visibly tighten. That feedback loop is the actual skill of running Google Ads in 2026: you are no longer writing every ad, you are training the system that writes them.

Bottom line

AI Brief is the steering wheel for AI Max. Left blank, you get generic automation and generic waste. Written with real constraints and iterated against search terms, it turns Google's AI into something that actually behaves like it works for your brand. Write the brief before you scale the budget.