For Tomas. Everything you need to introduce, qualify, and pitch ChatGPT ads to OO clients. Scroll, click into any audit, ready for sales calls.
The opportunity in one paragraph
OpenAI launched paid ads inside ChatGPT in February. Self-serve advertiser access opened in April. The minimum spend dropped from $250k to $50k commit, with $5k/month per campaign. Free + Go users see ads at the bottom of relevant answers (Plus/Pro do not). Most local businesses don't even know this exists yet. This is the first-mover window. We're moving on it now.
1What ChatGPT Ads is
OpenAI started showing sponsored placements at the bottom of ChatGPT answers in February 2026. They look like a small "Sponsored" card after the AI's normal response. Click takes the user out of ChatGPT to your landing page, just like a Google ad click.
Where ads appear
Bottom of answer
Always labelled Sponsored, visually separate from organic answer
Who sees them
Free + Go users
Plus / Pro / Enterprise / Edu = no ads
Targeting unit
Conversation prompts
Not keywords. 30-50 prompts per cluster
Pricing model
CPC live ($3-$5)
CPMs falling fast: $60 launch -> ~$25 nine weeks in
2How it works (mechanically)
The advertiser flow
Apply at openai.com/advertisers (waitlist)
Get approved (1-3 weeks based on current reports)
Set up advertiser account with OpenAI advertiser ID + payment method
Install OAIQ tracking SDK on the landing page (we do this for every client)
Build prompt clusters - 30-50 realistic prompts customers would type into ChatGPT
Write creative - 3 headlines x 3 bodies x 3 CTAs per cluster, "helpful tone, not promotional"
Launch + monitor with kill-switch on bad metrics (CPC/CPL/conversion-rate thresholds)
OAIQ - OpenAI's tracking system
OpenAI has its own pixel called the OAIQ SDK (Open AI Query). When someone clicks an ad, the click URL carries an oppref token. The SDK on the landing page reads it, stores in a 30-day cookie, and pings OpenAI back when the visitor converts (form, call, booking). It's like the Google or Meta pixel - just for ChatGPT ads.
The wedge: none of our 10 clients have this installed. None of their competitors do either. The first agency (us) to install OAIQ for our clients gets a 30-day attribution head start before the auction crowds.
3Limitations + restrictions
Money
Account commit
$50,000
Minimum total spend on the OpenAI account (down from $250k pilot)
Per campaign / mo
$5,000
Industry-quoted floor. Below this, signal too thin
Daily floor
$500
OpenAI's daily campaign minimum
Audience
Plus / Pro / Enterprise / Edu users see no ads. The audience is Free + Go tier ChatGPT users. For most local services this is fine (price-conscious customers exist in every income bracket). For premium-only / luxury / enterprise B2B targets, the free-tier audience may skew weaker.
Vertical exclusions (during current test phase)
Excluded
OO clients affected
Health (incl. dental clinical care, mental health)
Per industry reporting (ALM Corp), the first 90-180 days of any ChatGPT ads campaign produce imprecise measurement. Build reporting around a range of signals (OAIQ + GA4 + CallRail + form data), not a single conversion metric. Tell clients this upfront.
4How to qualify a client (4-axis scorecard)
Every readiness audit scores the client on 4 axes. 4/4 means clean GO. 3/4 with conditions. 2/4 means WAIT. 1/4 means probably no.
A. Vertical eligibility
Niche is on OpenAI's allowed list. Local services (movers/roofers/contractors), travel, ecom, education, professional services = allowed. Health, finance, crypto = excluded.
Pass = run readiness. Fail = stop, position AI-ranking organic instead.
B. ICP fit
Customers span income brackets (not premium-only). Buying decision is research-heavy. Avg ticket >= $500. Already converts paid traffic somewhere.
3+ yes = ICP fit. Below = flag risk in audit but may still proceed.
Client can commit $5k+/month per campaign. Avg ticket * close rate * margin > expected CPL after 3 scenarios (worst $5 CPC / mid $4 / best $3).
Worst-case ROI > 1x = GO. Below = WAIT or restructure.
Donal handles all of axis A, C, and D for you. The Tomas conversation focuses on axis B (does this client want more leads, what's their close rate, what's their avg ticket) - the rest you can lean on the audit deck for.
5Who it fits (industries + OO portfolio)
Industries already winning the beta (per Adventure PPC reporting):
Local services - movers, roofers, contractors, home services. This is OO's portfolio.
16.3x ROI, free pilot Month 1, OAIQ installs in Astro rebuild = no extra ask of Chris
Second
Soft Touch Moving
Tier-1 funder, weak organic = ADS-FIRST scenario, John wants better lead quality
Third
Bobcat Movers
5.0 review profile, Blake engaged, San Marcos market uncrowded
Why these and not others:
Elite + Moving Pro need LP form fixes first (worth doing, just adds a step)
Maestro's math is marginal until reprice closes 15 May
MW + ANC + Lankford are GO but each has one ahead-of-launch task
Howard / Soft Touch / Bobcat are clean
Plus we run onlineoptimisers.net in parallel as the eat-own-dogfood test. OO is the advertiser of record. The case study from running ads on ourselves becomes the strongest sales asset for the rest of the portfolio.
9Tomas sales playbook
The hook (use in opening message)
"ChatGPT just opened paid ad slots at the bottom of its answers. Most local [niche] businesses haven't even noticed yet. We checked your site - you'd likely qualify. Want me to send a 5-minute readiness audit?"
Recommended client ad spend: $5,000/month per campaign. Below this, statistical signal is too thin to optimise.
How to book a client onto this
Get prospect URL or pick from existing retainer list
Tell Donal: "Run /openai-ads-readiness on [URL]"
Donal returns audit deck link in 5 minutes
Send prospect: "Audit's done. Quick scroll. Recommendation at top."
If interested, ask Donal to run /openai-ads-vs-organic-decision for the decision memo
Send proposal (Donal generates from template)
Close on phone or email. 30-day notice contract.
Common objections + your answers
They say
You say
"How is this different from Google Ads?"
"Different audience, different intent. Google catches them while searching. ChatGPT catches them while researching. We can show you the prompt clusters where your customers are already asking ChatGPT the question your business answers."
"How do I know it'll work?"
"You don't, in Month 1. Neither does anyone. The first 90 days are imprecise across the industry. What you do know is the kill-switch pauses within an hour if metrics go bad, and the audit has shown you the math before you commit."
"Can you guarantee leads?"
"No, and any agency promising hard lead numbers in the first 90 days is making them up. What we guarantee is full transparency on every metric, the kill-switch on bad performance, and 30-day cancellation."
"Why $5k/mo minimum?"
"OpenAI's own daily floor is $500. Below $5k/mo statistical signal is too thin to optimise. We'd rather not run a campaign than burn through a $2k budget without learning anything."
"I'm already on Google Ads, why add this?"
"Different placement, different audience. Free-tier ChatGPT users number in the hundreds of millions. Most haven't been touched by ads in this surface yet. Early movers get cheaper clicks before the auction crowds."
What to flag to Donal immediately
Prospect is in an excluded vertical (don't waste time pitching)
Prospect's budget is below $5k/mo (math doesn't work)
Prospect already runs ChatGPT ads with another agency (we don't poach)
Prospect wants pay-per-lead instead of managed (Model A path - Donal scopes)
Anything that doesn't match this playbook (new edge case)