How to Run ChatGPT Ads: 12 Tactics for Better Results

We spent our own money running ads in OpenAI's Ads Manager Beta targeting agencies that need white-label PPC and SEO fulfillment. We documented what happened in our ChatGPT Ads field report. This piece is the follow-up: the tactical playbook we'd run in round two, combining what we learned with what other operators in the channel are seeing.

‍

A few notes before the tactics: performance in ChatGPT ads varies more by vertical and business model than almost any other paid channel. Borrowing someone else's CPL or ROAS as your benchmark will mislead you. The tactics below apply broadly, but the vertical section at the end covers where the biggest differences show up.

‍

Sharpen Creative and Copy

‍

ChatGPT ads land in a conversational environment, next to detailed, personalized answers. Creative that reads as generic or sales-first gets skipped; creative that mirrors the specificity of the conversation earns the click.

‍

Write Much Shorter Ads

‍

Anything on the tail of a description gets cut, and mobile dominates. Front-load the hook. Test putting your offer text directly in the 256×256 image. We saw competitors do this effectively (one ran a "free audit" line right in the creative). Note there's no video format yet, so static creative is the whole game right now.

‍

‍

Lead with Help, Not a Pitch

‍

People in ChatGPT are in learning mode. Nearly every competing ad we saw was sales-first ("hire our agency"). An offer that extends the conversation, like a free audit, a useful benchmark, a no-fluff comparison guide, or matching ad credit, earns the click. B2B buyers aren't signing today; give them a reason to take one small step.

‍

Get Specific and Quantified

‍

"Grow your business" is invisible next to ChatGPT's detailed answers. "Cut PPC fulfillment cost without hiring in-house" or naming the buyer ("built for boutique agencies") reads as relevant, not generic.

‍

Test Human Creative Over Logos

‍

A recurring report from the community: swapping a company logo for a real professional headshot or personal-brand-style image lifted CTR substantially. One advertiser cited a jump to around 5%. A conversational environment seems to reward a human face over a corporate mark. Worth an A/B test out of the gate.

‍

Rethink Targeting and Structure

‍

You can't see queries and you can't optimize to conversions yet, so the levers you do have (context hints, ad group structure, budget pacing) carry more weight than they would on a mature platform. Get these right and you shorten the learning curve.

‍

Simplify Your Context Hints

‍

This is the most useful thing other operators have surfaced, and it's counterintuitive. Several advertisers found that long, comprehensive context-hint blocks killed delivery; the system barely served their ads. Stripping them back to the essentials (business type, buyer intent, core use cases) fixed it. Think of the hint less like a keyword list and more like a short, clear GPT prompt: describe who you're for, why you matter, and the adjacent topics around your category, rather than cramming in every phrase you can think of. One advertiser put it well: cover the semantic understanding of the topic, not a pile of long-tail queries. Because location is still country-level only, your context hint is also how you approximate a service area.

‍

Run More Ad Groups and Creative Out of The Gate

‍

Because you can't see queries, the only way to learn what's working is structural testing: multiple intent-themed ad groups, multiple ad and image variants from day one.

‍

Use Daily Budgets, Not Lifetime

‍

Lifetime-budget pacing is reportedly not well optimized yet, and there's no ad-scheduling control. Daily budgets give you steadier, more predictable delivery until OpenAI ships scheduling.

‍

Fix the Landing page and Conversion Path

Traffic quality doesn't matter if the landing experience breaks the conversation. ChatGPT visitors arrive with unusually specific intent, and the page they land on either extends that intent or wastes it.

‍

Use the Conversational Ad Format When You Can

‍

The format that matters most for lead gen is the interactive "Ask ChatGPT about this ad" unit, where you pre-load product info, FAQs, and pricing and let ChatGPT field qualifying questions on your behalf. A prospect who has a real back-and-forth before they ever hit your site is dramatically more pre-qualified than a cold form fill. We ran the basic card; next time we'd pursue the conversational unit.

‍

Put a Chatbot on Your Landing Page

‍

This one surprised even the advertiser who reported it: ChatGPT traffic converted noticeably better when the landing page had a chatbot waiting. It makes sense in hindsight. Someone who just left a conversation is primed to continue one. If you're sending conversational traffic to a static form, you're forcing a gear change at the worst moment.

‍

Add Proxy Conversion Events

‍

Track engaged-visitor signals (scroll depth, key page actions) alongside the form fill. Even without bid optimization, it sharpens your quality read in GA4. Add a post-conversion "How did you hear about us?" survey to catch assisted conversions that click attribution misses, which matters for multi-touch B2B.

‍

Set Up Strategy and Measurement

‍

Two things separate serious tests from throwaway ones: coordinating with your organic AI presence, and funding the test long enough to learn from it.

‍

Show Up Organically in the Same Prompts

‍

Paid and organic AI presence reinforce each other, and you can be buying an ad next to an answer that doesn't mention you, or worse. Before bidding, run your target prompts 10 times and read what ChatGPT says about your category and your brand. If the organic answer is favorable, your ad has something to amplify. If it's not, fix that first. This is the new "own both the paid and organic result" play from early search, and it's where AEO/GEO work pays off.

‍

Fund It Like a Real Test

‍

For B2B lead gen, plan $3,000 to $5,000+ per month for 60 to 90 days to get a trustworthy read on cost-per-lead and lead quality. At minimum, triple our $500 (knowing you'll be bidding $10 to $12) to give yourself a shot at multiple leads in the door. Set the success metric for the first 90 days as "which conversation contexts produce quality leads," not "did we hit target CPL."

‍

Where Results Vary: Verticals and Business Models

‍

Performance in ChatGPT ads varies more than in almost any other paid channel, and borrowing someone else's CPL or ROAS as your benchmark will mislead you. Here's how the major categories break out.

‍

B2C and Ecommerce

‍

The link-to-product format makes this look like prime B2C real estate, but there's a credible counterargument that e-commerce is the segment that needs ChatGPT ads least. B2C brands already have Google Shopping, Amazon, and Google Ads as proven converting channels. Where it does work for B2C is micro-moment capture: being present at the exact moment someone is deciding ("high-protein dinner under 30 minutes"). Feedback loops are faster here, and a well-built B2C test can show meaningful conversion data in 30 to 45 days. Typical CPCs run lower, roughly $3 to $6 for product research.

‍

B2B and Considered Purchases

‍

This may be the stronger early fit. ChatGPT is increasingly where buyers do research and build shortlists before ever contacting a vendor. The catch is the cycle. B2B advertisers routinely judge the channel at two to three weeks, see no pipeline, and declare it dead, which is far too soon. Measure leading indicators (lead quality, demo requests, content engagement) early and pipeline later. Expect higher CPCs ($8 to $15 for complex B2B software) offset by stronger downstream conversion.
‍

Lead Magnets and Educational Offers

‍

Lowest CPCs (roughly $1.50 to $4), longer cycles. Good for filling the top of a nurture funnel.

‍

Restricted and Sensitive Verticals

‍

Health, mental health, and politics are excluded from targeting, and categories like crypto, financial trading, and some regulated industries are restricted. If that's you, factor it in before planning a test. Healthcare advertisers in the community report getting few or no impressions for exactly this reason.

‍

‍

What Other Advertisers Are Reporting

‍

The range of outcomes across the channel is the real headline. From advertisers comparing notes publicly, the results split sharply by execution, not just by vertical.

‍

Wins and Struggles in the Same Verticals

‍

One agency running ChatGPT ads in-house had form submissions coming in the same evening they launched, called conversion quality mid-to-bottom funnel at a much better cost than Google, and said they'd already closed a few new clients in the first few weeks.

‍

Another spent roughly $5,000 over three weeks in specialized home services and had the opposite experience. Their biggest complaint wasn't cost, it was visibility: no way to see how context targeting was affecting match quality, and so no idea what to change.

‍

Tellingly, home services shows up on both the "crushed it" and "struggled" lists depending on who's talking. That's the point: the variable isn't just the vertical, it's how well the targeting and creative are executed within it.

‍

CPCs Swing Hard By Category

‍

One advertiser in B2B consulting couldn't even enter the auction at a $4.50 bid. Another in home goods retail was told $4 should win clicks. A B2B software advertiser reported $4 to $5 CPCs with roughly half of clicks registering in GA4.

‍

Reported winning niches have included home services, car rental, international real estate, and some legal (personal injury, DWI). Reported strugglers have included criminal defense, tours, employment law, early-stage SaaS, and dental.

‍

The Early-mover Discount is Eroding

‍

Advertisers who saw CPAs three to four times cheaper than Google earlier in the year report that edge compressing toward roughly two times as more competitors pile in. At least one watched CPMs fall from roughly $60 to $25 over nine weeks as inventory opened up. The cheap window is real, and closing.

‍

Set Your Own Baseline

‍

A B2B SaaS result tells you almost nothing about a DTC apparel result, and a number you read in a case study (or a Reddit thread) tells you almost nothing about your vertical until you've run it yourself with a proper media planning approach.

‍

Plan a ChatGPT Ads Test Worth Running

‍

The channel is still young enough that the operators who learn the conversational creative game now, while CPCs are climbing but competition is still relatively thin, will have a real head start when the tooling matures. The tactics above are where we'd start. What actually moves the needle for your account depends on your vertical, your funnel, and how patiently you can fund a real test.

‍

That's the work our paid media team does every day: running and measuring tests in emerging channels, translating what works into a plan tied to your pipeline. If you're planning a ChatGPT ads test and want a partner who's already learned what the dashboard won't teach you, let's talk.

‍