The Briefing
1. Google rebuilt its entire ad stack around Gemini. At Google Marketing Live on May 20, Google announced Ask Advisor — a unified AI assistant that sits across Google Ads, Analytics, Merchant Center, and Google Marketing Platform at once. Also new: Conversational Discovery Ads that write their own creative based on what a user is actually typing, and a Universal Cart that lets buyers add products from Search, YouTube, Gmail, and the Gemini app. If you're running Google Ads, log in this week and learn what changed. It's not a "watch this space" moment — it's a "this is already live" moment.
2. ChatGPT ads are now open to any business. OpenAI launched a self-serve Ads Manager and killed the $50,000 minimum spend that locked out most advertisers. Early data from brands already running: click-through rates roughly 3x higher than traditional search, and conversion rates approaching 2x in several categories. It's early. It's still expensive to test. But those numbers are hard to ignore for anyone running paid traffic.
3. Meta just doubled your retargeting window — whether you asked for it or not. Starting May 18, Meta automatically expanded existing purchase-event custom audiences from 180 days to 730 days. That's nearly two years of buyer data available for retargeting. Great news for high-ticket products and seasonal businesses. But if your retargeting strategy depends on recency — buyers in the last 30 or 60 days — go check your Audience Manager now. Some audiences silently expanded in ways you didn't intend.
4. Adobe Firefly can now run full creative workflows from a single brief. Adobe's new agentic AI takes one brief and executes across Photoshop, Illustrator, and Premiere — no switching apps, no repeating yourself. A lighter version works inside Claude directly. For anyone producing video ads, social creative, or landing page assets, the time-cost of production just dropped again.
The Build
This week I built a focus group.
Not a real one. No Zoom call with five strangers, no incentives, no moderator. A simulated one — a panel of ideal clients that lives inside Claude as its own Project, ready to review anything I put in front of it. Copy. Offers. Landing pages. Emails. Subject lines. Positioning angles. All of it.
The idea came from Justin Brooke (@IMJustinBrooke on X). His setup was slightly different to mine. But I took his core concept — deeply built client avatars that Claude can embody — and turned it into a dedicated Claude Project so the panel is always available, always in character, and never has to be rebuilt from scratch.
Before I set it up, I went deep on the avatar creation. This part matters more than the setup itself. A shallow avatar gives you shallow feedback. I spent real time on the Ideal Customer Profile — not just demographics, but psychographics: what they believe, what they fear, what they've tried before, what objections live rent-free in their head, and what outcome they're quietly desperate for. The more specific the avatar, the more useful the feedback.
Here's the core prompt I used to build the panel:
You are a panel of 5 ideal clients for [your business/niche description]. Each panel member has a distinct but realistic profile: Member 1 — [Name, age]: [Job/role. Primary pain. Key desire. Core objection to buying. One belief that shapes how they see the world.] Member 2 — [Name, age]: [Same format] Member 3 — [Name, age]: [Same format] Member 4 — [Name, age]: [Same format] Member 5 — [Name, age]: [Same format] When I share copy, an offer, a landing page, or an email with you, respond as each panel member would — in their own voice, from their own perspective. Give me honest reactions: - What lands? - What confuses them? - What triggers resistance or doubt? - What makes them want to keep reading or take action? - What would make them close the tab? Do not be polite. Do not encourage me. Be the client. Each member should respond individually. Their feedback will be used to improve the work, so honesty matters more than encouragement. Confirm you understand your role as each panel member before we begin.
Once the panel is set up inside its own Claude Project, you feed it anything and ask for feedback. I sent it a landing page headline I'd been debating. Within seconds I had five distinct reactions — not generic "this is good" or "try being more specific" feedback, but responses in character.
One panel member said the headline felt like it was written for someone further along in the journey than she was. Another said the word choice felt like it came from a consultant, not someone who'd actually done the work.
Both were right. Both were things I probably wouldn't have caught myself.
What struck me was the quality of the learning. This isn't just copy editing. It's market intelligence.
The panel tells you how your ideal client actually processes what you've written — what associations they make, what assumptions they bring to the page, what meaning they read between the lines.
You start to understand not just what to change, but why it wasn't working.
It reminded me of what Justin said about the setup: the goal isn't to get approval from the panel. The goal is to understand how real people in your market think. Once you understand that, every piece of marketing you write gets sharper.
What worked: The depth of feedback was genuinely useful, not performative. Once the avatars were built properly — specific beliefs, specific objections, specific language patterns — the panel responded like real people, not like an AI trying to be helpful. The "do not be polite" instruction was critical. It changed the entire tone of the output.
Setup was faster than I expected. The time investment is almost entirely in the avatar creation, not the technical setup. That part took maybe 20 minutes and the Project has been running ever since.
What didn't work: My first pass at the avatars was too generic. I leaned on broad demographic details and got broad feedback. It wasn't until I added specific beliefs ("she thinks most business coaches are just selling hope"), specific prior attempts ("tried a group program before, felt unsupported"), and specific language quirks that the panel started giving me feedback I couldn't have given myself.
Garbage in, garbage out — this applies more here than almost anywhere else.
The Play
This one is for the person working alone — by themselves — wondering how to get real feedback on their ideas.
You know the problem. You write a sales page or an email and then you read it back to yourself. You've been staring at it so long you can't see it anymore. You know what you meant to say, so that's what you read, even if it's not what's on the page. You need a second opinion. Preferably five. From people who actually represent your market.
Here's how to build your panel today.
Step 1: Build your ICP first. Before you create any avatars, get specific on who your ideal client actually is. Go beyond job title and income.
What do they believe that's holding them back? What have they already tried? What's the sentence they say to themselves at 11pm when they can't sleep? What objection is so ingrained they'd never say it out loud on a sales call?
Answer those questions in writing before you build a single avatar. The panel is only as sharp as the input you give it.
Step 2: Create the Claude Project. In Claude, create a new Project (not a regular conversation — a Project, so the panel persists). Use the prompt from The Build section above. Build 10-14 panel members. Give each a name, an age, a distinct pain point, a distinct objection, and a distinct belief. The more they differ from each other, the more useful the panel becomes.
Step 3: Feed it your work. Paste in whatever you want reviewed — a headline, an email, a landing page, an offer structure, a subject line.
Then ask: "How does each panel member react to this? What works, what doesn't, and what would they do next?"
Run this follow-up prompt whenever the feedback feels too surface-level:
Go deeper. For each panel member who had a negative or hesitant reaction — what is the specific belief or past experience that's driving that reaction? What would I need to say or show to address it? Give me their inner monologue, not just their surface response.
This is where the real learning is. Not the headline tweak. The underlying reason the headline didn't land.
If you're a solopreneur, a copywriter, a marketer who reviews their own work more than anyone else's — this tool will change how you write. Not because AI is smarter than your market. But because it gives you enough distance from your own work to actually see it.
Try this today. Build the avatars. Give them a name, a real pain, a real objection, and one belief that shapes how they see your market.
Then take a piece of copy you've been unsure about and ask the panel how they'd react. You'll know within 60 seconds whether this is going to become part of your process.
The Persona Builder
Copy this. Drop it into a new Claude Project. Your focus group panel is ready in under 20 minutes.
Step 1 — Answer these before you build anything. Write your answers out in plain language. This is for you, not for Claude. The quality of your panel depends entirely on how well you know these answers.
About your market: Who is the specific person you're reaching? What's their #1 problem right now? What have they already tried that didn't work? What do they believe is the reason things haven't worked (their theory, not the real one)? What outcome are they quietly desperate for — the thing they'd never say in a survey?
About their objections: What's the first thing they think when they see an offer like yours? What's the belief that makes them hesitate even when they know they should act? What would a trusted friend have to say to get them to finally move?
About their language: What exact words do they use to describe their problem? What insider phrases do they use? What do they call themselves?
Step 2 — Build your panel. Paste this into a new Claude Project and replace every bracket with your specifics.
You are a panel of [NUMBER] ideal clients for [YOUR BUSINESS — what you sell and who you sell it to in one sentence]. Your job is to give honest, in-character feedback on marketing materials I share with you. Each panel member has a distinct profile, voice, and set of objections. --- PANEL MEMBER 1 — [FIRST NAME], [AGE] Role/situation: [e.g. "Runs a 3-person accounting firm, 8 years in, revenue has plateaued"] Primary pain: [e.g. "Working 60-hour weeks and still not profitable enough to hire"] Core desire: [e.g. "Wants to work less, earn more, and feel like the business works for her — not the other way around"] Key objection: [e.g. "Sceptical of coaches — tried one before, felt like she paid for homework she didn't have time to do"] Belief that drives her: [e.g. "Thinks the answer is probably a systems problem but doesn't know where to start"] Language she uses: [e.g. "Overwhelmed, spinning my wheels, I should be further along by now"] --- PANEL MEMBER 2 — [FIRST NAME], [AGE] Role/situation: Primary pain: Core desire: Key objection: Belief that drives them: Language they use: --- [Repeat for each panel member — build 4 to 6 total] --- INSTRUCTIONS FOR ALL PANEL MEMBERS: When I share copy, an offer, a landing page, an email, a subject line, or any other marketing material, each panel member responds individually — in their own voice, from their own perspective. For each piece of material, give me: - First reaction (gut response in 1–2 sentences) - What lands (what resonates, what feels true) - What creates resistance or doubt (what makes them pause, pull back, or disengage) - What confuses them (anything unclear or that requires re-reading) - What they'd do next (click, keep reading, close the tab — and why) Do not be polite. Do not tell me what I want to hear. Respond exactly as these people would — with their real hesitations, their real language, and their real decision-making process. Each member responds separately. Label each response with the member's name. Confirm you understand your panel role before we begin.
Step 3 — Run feedback. Paste in any piece of marketing and ask: "Panel — here's [describe what it is]. How does each of you react?"
Step 4 — Go deeper when feedback feels surface-level.
Go deeper on [PANEL MEMBER NAME]'s reaction. What is the specific belief or past experience driving that response? What would I need to say, show, or prove to move them past it? Give me their inner monologue — not just the surface reaction.
Step 5 — Test your rewrites.
I've rewritten the [headline / email / offer / page]. Here's the new version: [PASTE REWRITE] Same question — how does each panel member react now? What changed? What still isn't landing?
Save this Project in Claude so the panel persists across sessions. You won't need to rebuild it. Run new copy past the panel before you publish. Run existing copy past it when conversion drops. Run your offers past it before you change the price.
The goal isn't approval. The goal is to understand how your market actually thinks.
Download ai-focus-group-persona-builder.md
Growth Partner for founders stuck between $1M and $20M. I find the money you’re sitting on and build the machine that collects it. $1B+ in client results.