When your ad becomes a chat, your offer has to survive questions

This sounds like a small ads update. It is not.

Google is testing ads in AI Mode and Search where Gemini can explain why a product or service fits a question. In the same package, Google describes Conversational Discovery ads, Highlighted Answers, AI-powered Shopping ads, and Business Agent for Leads, where someone can click "Chat" inside an ad instead of only filling out a static form.

For a small business, that changes the prep work. Your ad will not only need a headline, an image, and a landing page. It will need to survive follow-up questions.

Source: New ad formats built with Gemini coming to Google Search

Google Marketing Live 2026 had a clear pattern. Gemini is moving deeper into ads, shopping, Merchant Center, Analytics, and creative workflows. Google describes Ask Advisor as an AI collaborator across Google Ads, Google Analytics, Merchant Center, and Google Marketing Platform. That matters if you work with paid channels. But for Hammer Automation's audience, the interesting part is not the button Google adds next.

The interesting part is the requirement before the button: your offer has to be clear enough for AI to explain it without guessing.

Source: Google Marketing Live 2026: News and announcements

What changes when the ad becomes a conversation?

A normal search ad usually answers a short keyword. An AI-powered search experience starts when the customer writes like a human:

"We need a calm place for a small team kickoff near Gothenburg, preferably with vegetarian food and the option to book this week."

"Which accounting firm fits if I run a sole proprietorship and want to stop chasing receipts?"

"Which AI training can a school staff finish during one in-service day without it becoming a tech show?"

Google says Gemini can evaluate and synthesize information about the advertiser's product or service and display that context alongside the advertiser's creative. That makes the ad feel less like a poster and more like a short piece of advice.

Good. Also revealing.

If the website is vague, if the price is hidden, if the conditions live only in someone's head, AI has too little to work with. It can still produce a polished answer, but the answer will be fragile. For small teams, this is a useful reminder: AI marketing does not start in the campaign tool. It starts in what you know about the offer.

Start with an offer brief, not an ad

An offer brief is a short working document that describes what you sell, who it fits, what is included, what is not included, how pricing works, which questions customers usually ask, and when a human should take over.

It does not have to be pretty. It has to be clear.

Think about a café selling catering, a consultant selling AI workshops, a school promoting an open house, or a small ecommerce store with products people compare before buying. All of them can spend money on ads. But if AI is going to answer customer questions, it needs to know the difference between "can be delivered Friday" and "can sometimes be delivered Friday if someone has time".

That is where a lot of small business marketing wins or loses. Not in the final slogan. In the details that never got written down.

45 minutes: make one offer ready for AI questions

Pick one offer you already sell. Not the whole company. One package, product, service, or recurring booking.

Minute 0-10: write the offer without sales language

Answer briefly:

• What does the customer get?
• What does it cost, or how is the price calculated?
• How quickly can the customer get it?
• What limitations exist?
• Which customers is it not right for?

Minute 10-20: collect real customer questions

Pull questions from email, DMs, calls, quote requests, support tickets, and the Google searches you already see. Write them the way the customer would write them, not the way you wish they asked.

Minute 20-30: label every answer with evidence

Add where the answer comes from:

• a product page
• a price list
• a contract
• a delivery policy
• a calendar rule
• an internal routine
• a human judgment call

If the answer only lives in someone's head, write "needs documentation". That is not a failure. That is the point of the exercise.

Minute 30-40: decide the stop points

AI can answer common questions. It should not promise a special price, overbook capacity, give legal advice, or confirm something that needs human review. Write down when the answer should become: "I can help prepare this, but a person needs to confirm it."

Minute 40-45: run one test question

Ask your AI assistant a difficult but realistic question and tell it to answer only from the brief. If information is missing, it should say exactly what is missing.

Copy this prompt: test your offer before Google does

Use this in ChatGPT, Gemini, Claude, or another tool. Replace the bracketed parts.

You are my review assistant for an offer that may be described in ads, AI search, customer email, and chat.

Goal: find unclear points before a customer or AI agent misunderstands us.

Offer:
[paste the offer brief]

Customer type:
[example: local restaurant customer, parent, ecommerce buyer, school principal, consulting client]

Do this:
1. Write ten realistic customer questions in natural language.
2. Answer only from the brief. Do not guess.
3. Label every answer as: clear, unclear, or needs human review.
4. List the missing information required before the answer is safe to use in an ad, website, or chat.
5. Suggest three landing-page improvements that make the offer easier for both people and AI systems to understand.
6. Suggest one short stop rule: when should AI hand the question to a human?

Write practically. No campaign slogans.

This is not a prompt for making the ads more entertaining. It is a prompt for finding the holes before a customer does.

Three small workflows to try this week

Catering inquiry to quote brief

When someone asks about catering, AI can read the offer brief, pull out number of guests, date, food preferences, and delivery location, then draft a quote brief. A person approves price and capacity before anything is sent.

Course interest to the right training path

A school or organization can let AI compare the customer's needs against three described training packages: introduction, practical workshop, and deeper follow-up. AI suggests a path and questions to ask, but booking and tailoring stay with the responsible person.

Product questions to a better product page

An ecommerce store can feed in ten questions customers often ask. AI marks which answers actually exist on the product page and which require new text, better images, measurements, return rules, or delivery times.

Integrate safely without making a mess

When AI connects to ads, CRM, calendars, or ecommerce, keep the first integration narrow. Give AI read access to the right documents first. Use environment variables or a secret manager for keys, not passwords in chat. Use scoped API keys where possible. Let AI create drafts, summaries, and briefs, but require approval before a price, booking, discount, or promise reaches the customer. Save run logs. Redact unnecessary personal data in outputs.

That is not fear. It is normal work management, applied to AI.

For Hammer customers, the path often looks like this: Mindset Forge to choose the right offer and customer journey, Tool Forge to build the small integration, and Skill Forge to train the team, so the routine is used the same way each week.

Frequently asked questions

What is Business Agent for Leads?

Google describes Business Agent for Leads as a Gemini-built agent inside a Search ad. A person can click "Chat" and ask questions based on the advertiser's website, instead of only filling out a form.

Do small businesses need to use Google's new ad formats right away?

No. The useful move this week is not switching on a new format. It is making one offer clear enough for customer questions, AI search, and future chat flows.

What is AI Mode in Search?

AI Mode is Google's more conversational search experience where a person can ask longer questions and get summarized answers, recommendations, and follow-up questions.

What should we document first?

Start with one offer that already gets questions: price, delivery, target customer, limits, common objections, and when a human should take over.

The practical shift

AI in ads does not automatically make marketing better. It makes good source material more valuable.

If your offer is clear, AI can help the customer get to the right question, product, or booking faster. If the offer is fuzzy, AI becomes a faster route to the same confusion.

So start there. Pick one offer. Write the brief. Test ten questions. Close the gaps. Then you are better prepared whether the next customer arrives through Google, email, chat, or a person who simply wants a straight answer.