We never like to recommend, or try to implement, a new form of online advertising, until we’ve tried/tested it with our own business. That way we can understand the pros and cons of running advertising on a new advertising vehicle, plus all the “inputs” we’ll need to provide to the tool to launch the campaign. That makes it easy to know what questions we need to ask of our clients related to a new ad campaign launch on a vehicle that is new to them, or what we need their approval on.
We waited to try this new vehicle until they offered pay-per-click (PPC) vs. just pay-per-impression (CPM) advertising because, quite frankly, most of clients are small-to-mid-sized businesses (SMBs) and nonprofits and they have far smaller marketing/advertising budgets than large firms and need their ads to really work for them and generate leads and sales vs. just awareness-generation!
As of the original publish date of this blog post, It’s been about 1.5 weeks since we rolled out ads for our own business, Results Communications & Research, on the ChatGPT LLM-powered (large-language-model-powered) answer engine/AI search platform. We’ve documented below — and will continue to document — our 2026 journey with this new form of online advertising. So, be sure to check back!
- Days 1 & 2 (June 11 and June 12) – set up our ad and associated targeting for our campaign at ads.openai.com (ChatGPT is owned by OpenAI) — it was not a large lift at all! Needed to determine and document in OpenAI’s ad tool how much we were initially wanting to spend for a click (we initially put this at $3.50) and our daily budget ($25), plus in what geographies (Providence RI and Boston MA) we wanted the ads to appear. And, also provide context hints (OpenAI instructions explained this exercise this way: “describe the conversations, topics, or keywords where your products or services may be relevant; these hints guide matching, but aren’t exact-match targeting rules.” We also documented the images and content we wanted to use in our ads. We are testing two ads — one with longer and one with shorter headlines — but to keep something consistent, so we’d feel comfortable drawing conclusions about our headlines, we used the same image in both ads. We also had to share with the ad platform what page of our website we wanted ad clickers to land on (we created a brand new page on our WordPress website for such purposes — which is an online advertising best practice).
- Day 2 (June 12) – launched our campaign by activating it in our OpenAI Ads dashboard.
- Day 5 or 6 (June 15 or 16) – of course, being the digital marketing and analytics geeks that we are, we checked our advertising dashboard at openai.com countless times to see if the ads were being served up and generating impressions and/or clicks. Seeing zero movement on either the impressions or clicks front, we decided to expand our geography to NYC and our daily budget to $35 and what we were willing to spend for a click to $5.00, in case what we were willing to pay for a click (and the competition to have ads shown by competitors in the same geography for the same types of prompts) was causing the ads to not be shown or we needed a bigger geography to have enough opportunities for our ads to be shown to relevant “prompters” on ChatGPT.
- Day 7 (June 17) – we saw in our advertising dashboard that our ChatGPT ads were starting to accrue impressions, and therefore, were being served up, but not to any great degree.
- Day 8 (June 18) – we increased what we were willing to pay for a click to $7.50, and also added in the whole state of New Hampshire to our geographic targeting.
- Day 9 (June 19) – seeing an alert in the OpenAI ads platform that we were often being outbid by competitors because of what we were willing to pay for a click, we increased our click bid to $10.00 from $7.50 – our first 2 clicks were achieved somewhere between the 18th and 19th, and at that point, our ads had a 0.66% click-thru rate (meaning less than 1% of individuals having our ad served up to them, clicked on it — not surprising for reasons outlined in the next bullet.)
- Day 12 (June 22) – we just checked this morning and the increase in budget and what we are willing to spend for an ad click, plus the expanded geography seems to be paying off! As of this morning, our ChatGPT ads have generated 2,762 impressions (been served up that number of times), 13 clicks (a 0.47% click-thru rate), and an average cost-per-click of $6.44. We aren’t surprised by any of this. We had seen competitors running ChatGPT ads in our space when we were entering relevant prompts into ChatGPT, so we knew as a business-to-business advertiser we were more likely to have to pay more for a click than a business-to-consumer advertiser might. We also know that ChatGPT is a different animal than Google Search ads — the new form of ads discussed in this post is served up to individuals entering prompts into ChatGPT that might indicate they’d be interested in our services, but it’s not the same as someone entering “keywords” into Google and then Google serving up ads to those individuals because an advertiser provided those keywords as ones they wanted their ads to be served up for. We changed our daily budget to $50.00 today — we should have increased the daily budget a few days back in keeping with our increasing the amount we were willing to pay for a click! We also paused the shorter of our two ads since the longer ad was performing far better. To-date, we have not generated any leads — in the form of phone calls, e-mails, or appointment set-up — from these ads. We’re not discouraged by this at all, since, when it comes to a business engaging a firm for B2B services — or even reaching out to them — there can be a long lead time!
- Day 17 (June 27) – we paused our ads today. Not because they weren’t generating impressions or clicks, but simply because we are a small marketing agency/SEO company who deliberately keeps our fees/rates very reasonable so that small businesses and nonprofits can afford to work with us. Going into this pilot, we planned to only spend about $300, but ended up letting the ads run until ChatGPT ads had spent $364.12, based on our daily budget and what we were willing to spend per click.
How Successful Was Our ChatGTP Ads Campaign?
Because businesses often take some time before purchasing business services, and therefore, like to research a number of companies before they decide to reach out to the one organization or the few organizations whom they are most interested in, we have yet to receive any inquiries from our ads, but still hope to get some inquiries from those who have clicked on the ads. At least we’ve generated some “awareness” among appropriate parties. We’ve also found in our many years of running online ads for our own business, that often, fellow marketing agencies/competitors click on our ads to see what we are offering and what we are all about. We did ensure our landing page for our ad clickers adhered to best practices, such as demonstrating why we are experts, including “social proof,” therefore, a review or testimonial, and much more!
Ultimately our ChatGPT ads yielded:
- 11,321 impressions (this number of sets of eyeballs viewed our ads as part of entering an appropriate prompt into ChatGPT)
- 46 clicks
- 0.41% CTR (we aren’t concerned or scared off by this low click-thru rate — we’ve seen similar with Reddit and nextdoor.com ads
- Average cost-per-click (CPC) of $7.92 – not horrible for a B2B (business-to-business) lead
As we recommended at the top of this blog post, keep coming back to learn more about our ChatGPT advertising journey! We’re all in this together when it comes to understanding this particular form of new advertising and how effective it will be in generating leads and sales, not just awareness, and if ROI makes sense for advertisers like our clients who need their ad $$ to work for them pretty quickly.











