AI-Generated Ads - Worth the Gamble?

24fire

One of my long-term clients, a remodeler, recently asked me to churn out some Facebook slop for their page. The conversation quickly morphed into him wanting to throw budget at ads generated by some AI platform - video or image, didn't matter. Our regular campaigns haven't been pulling their weight lately, and he's in his forties, fully bought into the whole ChatGPT ecosystem for his business.

I've dabbled with AI images for his website before, but never for an actual ad. The thought of burning through his budget on something I can't control makes me uneasy. What's the real-world experience here? I've seen people claim loads of AI ads work just fine - it's all about the tool and the prompt. But that feels like saying a good paintbrush guarantees a masterpiece.

Others have pointed out that if I'm not deep into the ad game already, I'm setting myself up for failure. Professional tools are almost mandatory, and you need to understand the specific triggers that make homeowners care about remodels in your area. Otherwise, you're just throwing money at Meta's visual abyss.

So the question isn't really 'should I try AI ads?' - it's more about whether the process of prompt engineering and platform selection can reliably replace the human intuition behind a winning creative. I've seen too many startups burn cash on automated ad sets that look slick but convert like a lead balloon. Anyone here cracked that code without losing their shirt?

trafficcop

Oh, another "hot topic" that's been done to death - but fine, I'll bite. AI ads are like that intern who occasionally nails a brief but mostly just churns out generic drivel. Handy for cheap top-of-funnel testing, sure. but slap a real budget behind it and watch it crash harder than my patience with Meta Business Suite.

For a remodeler? Splits test between that AI nonsense and your own stuff. Don't go full hype mode. Keep the client happy and the account out of the ditch

prpro

AI photos can work if you have a good eye for what looks like AI. Like a really good eye, because the moment people recognize it as AI, you are going to piss people off.

Also, I have never seen a good AI human model in video form. Only photos. Anyone thinking AI videos of humans look acceptable right now, are likely boomers who don’t have an eye for this shit yet. I would not use videos yet.

Bottom line, it has to be done tastefully and cannot substitute creating ads using actual creativity.

I see a ton of AI ads a day, and most of them are completely shit and the comments are full of people roasting them. So tread lightly.

coopplayer

Tested for a dental client via Banana 2. The client had shite creatives so I just went from scratch. Best performing creatives to date. Only did images. Haven’t tested videos yet.

nightowl

No offence, but the way you're framing the question tells me you're still paddling in the shallow end of paid media.

First thing: for professional-grade video ads you're going to need a proper tool - don't mess about with free or experimental stuff. You can go to the source (like Runway or Pika) or a provider that bundles pricing, but you should do your own research. For me, it's a bit like picking a freight forwarder: the margin is in the middleman if you know where to look.

Also, you haven't mentioned how many variations you're willing to test - A/B, different hooks, maybe even different ad formats. Without that number it's impossible to give targeted advice.

The bigger question, though: do you actually understand what draws people to remodels in your local area? Are they looking for speed, cost, or a specific aesthetic? If the answer is fuzzy, your ads will miss the mark and you'll burn budget fast. On Meta especially, visual impact is everything - you can't fake relevance with a generic video and a prayer. Choose your budget allocation like you'd allocate capital in a downturn: defensively.

chatbox

From my side of the fence, the AI part isn't the problem per se. We've run a few tests with AI-generated creative on smaller, local service campaigns and it pulled decent numbers - especially when speed and volume trump a polished brand look. The real issue I've seen is people jumping on the tool before diagnosing why their existing campaigns slowed down. If lead quality or conversion rates were already dropping, AI ads can temporarily mask that with fresh creatives and a spike in engagement, but it doesn't always translate to better enquiries.

My approach would be to test it in a controlled way with a separate budget bucket rather than replacing your current setup outright. That way you can compare apples to apples in whatever analytics platform you're using - GTM plus GSC or Ahrefs gives you the granularity to see if the AI stuff actually changes on-site behaviour or just inflates vanity metrics.

One thing I'd flag: the AI image/video tools are good enough now that most users won't care unless the ad looks weird or overcooked. Keep an eye on the click-through rates and conversion path data, and use a separate campaign tag so you don't contaminate your existing performance data. If you're running on Meta, set up a dedicated ad set with a defined budget and let it run for at least two weeks before making any calls.

localpack

Honestly, I'd step back and ask what problem you're actually trying to solve here.

If you're chasing better results and think AI will just magic them up... that's a deeper issue with your offer or targeting.

If you reckon AI can create something truly compelling that stock photos or real footage can't touch - maybe. But 9 times out of 10, a full AI video screams "uncanny valley" and tanks trust. I've seen it crater CTR on high-ticket services especially.

Are you planning to use AI just for generating variants of a static image, or are we talking full-blown generated videos? because if it's the latter, ask yourself: would you trust a remodeller whose "after" shots look like they came out of Midjourney?

Run a simple A/B test - one control with real assets, one AI-heavy variant. get the data. That'll answer more than any theory session will. 🧠

nightowl

AI creatives can work for cheap testing-think of them as a quick sketch to validate a concept before commissioning the real art. I've seen them hold up in local service campaigns, but on a saturated digital shelf like Amazon, the generic feel accelerates fatigue. Best to pit them against your winning assets on a small budget first, never swap the proven for the convenient.

bouncekiller

AI ads? Sure, they can work - but the way you're asking tells me you'll just torch your budget

24fire

broad targeting these days is essentially Google's AI overview in disguise-you're already running AI ads whether you realise it or not. if you want to experiment on purpose, ChatGPT's ad platform is worth a look. not perfect, but it's the closest thing to letting an algorithm write and optimise its own copy

ranktracker

Honestly I think people are looking at this the wrong way when they ask “do AI ads work?”

The better question is: what part is actually AI here? Because there’s a huge difference between:

a) using AI to crank out 30 creative variations fast vs
b) running fully fake remodel videos with plastic-looking kitchens and weird AI humans

Those are not the same thing at all

From what I’ve seen, AI stuff absolutely can improve performance at the top of funnel. Especially on Meta where "novelty" still matters a lot. Weird compositions, exaggerated lighting, unusual hooks… they stop scrolls. Sometimes better than polished agency creatives honestly

But here’s a catch nobody talks about enough: higher CTR does NOT automatically mean better remodel leads

In remodeling specifically, trust is doing like 70% of the work. People are about to spend $20k-$100k letting actual strangers into their house. So the second the ad feels fake or “off”, you create friction you may never see inside Ads Manager

That’s why I think the middle-ground approach is the smart one right now:

use AI for concept testing
hooks
background expansion
thumbnail ideas
angle variations

but keep the actual projects, people, before/afters, trucks, crews, etc real

because local service businesses are different from ecommerce. If I see an AI-generated hoodie ad, whatever. If I see an AI-generated kitchen remodel, my first thought is “wait… did these guys even do this project?”

Also something I learned the hard way: AI creatives can accidentally mask bigger problems for a while. Fresh creative gives the account a temporary boost, so clients think “the AI fixed it”, when really the offer, follow-up, or lead quality issue is still sitting there underneath

So yeah I wouldn’t avoid it completely at all. I’d just treat it more like a creative acceleration tool than some magical replacement for actual marketing instincts

And personally… I still wouldn’t touch full AI humans/video for remodeling yet. Consumers are WAY better now at spotting it than marketers think lol

localpack

No and Yes is the right answer here, honestly. 🎯

I've been spending serious money on ads for years, and I've tested AI tools across dozens of accounts. Here's the framework I use:

AI for creative copy / image generation? Yes, use it. Speeds things up massively.
AI for targeting / bidding / keyword selection? Hell no, especially if you're working with low search volume terms. the moment you let it loose on thin data, it burns budget chasing phantom intent.

Someone in a thread recently said "don't trust AI, use it but check the work." That's literally my policy now. Generate the angles, but validate every single assumption. The algorithm isn't smart enough to know your customer like you do - especially when the data pool is small.

bottom line: treat AI like an intern that works fast but needs constant supervision. You wouldn't let an intern set your bidding strategy on a 100-keyword campaign, would you? Same rule applies.

trendspotter

AI ads are absolutely worth it when you treat them with the same creative discipline you'd give a photoshoot. i've got a client running a whole campaign built on AI-generated assets and the numbers are frankly ridiculous. the key is avoiding that horrible uncanny valley effect - you know, the one where it looks like an AI tried to pass off as a real human testimonial and your brain screams "fake." Instant death for brand trust.

for static images, you can produce results so hyper-realistic most people will swear they were shot in a proper studio. And for video, lean into something that doesn't pretend to be real - fun animated explainers or stylised motion graphics. just don't do a "realistic" human face talking at the camera. that's still a dead giveaway, and it creeps people out. A year from now that might change, but right now humans are still decent at spotting a digital imposter.

no need to worry about "wasting budget" either - you can churn out dozens of variations for pennies, test fast, and kill what's not working. The real question is whether you have the creative vision and strategic muscle to make ads that actually feel on-brand and compelling. only you know if you can pull that off

bouncekiller

Been running AI-generated creatives for ecom clients for a few months now. Honest take:

The statics are ready. AI-generated product images and ad creatives are at the point where your audience genuinely can't tell the difference. For a remodeler, you could generate before/after style visuals, project showcase cards, seasonal promo creatives all from a few job site photos. Quality is there.

Video is getting there but not fully cooked. AI avatar/UGC-style videos still have the uncanny valley problem. What works better right now is templated video real photos with animated text, transitions, voiceover. Looks professional, doesn't look "AI generated."

The budget question: don't think of it as "AI ads vs real ads." Think of it as "test 10 creatives for $50 instead of 2 creatives for $500." The AI part just gives you volume to test faster. Your targeting, offer, and landing page still do the heavy lifting.

Start small. Take 3-4 of his best project photos, generate a batch of creatives, run them at $10/day against his best performing manual ads. Let the data decide, not the vibe.

ppcnerd

I'd point anyone to that bagel shop article. It's a cautionary tale: AI can generate ads, but it can't generate trust. Worth trying if you keep human oversight - the macroeconomic lesson is that shortcuts often backfire.

trafficcop

Oh honey, "tons of AI ads that work just fine" is like saying there are tons of decent canapés at a conference-technically true, but you've still got to dodge the soggy ones. The tool and prompt matter, sure, but mostly it's about how much effort you're willing to shovel into the prompt abyss. I've seen so many "AI-generated" ads that scream "I typed 'fun B2B vibe' and hit enter." Spoiler: they didn't work. Unless your definition of "winner" is getting roasted by your own CMO.

cpaoptimizer

My take: treat AI ads as a cheap testing layer, not your whole campaign. Grab a few UGC-style hooks from a tool like UnrealUGC and only scale the ones that outperform your existing creatives.

logan-lifecycle

Treat AI ads as a separate branch in your customer journey map. Start by isolating a small, low-risk segment-maybe a lookalike audience or a specific trigger node-and run them as a side experiment alongside your main flows. That way you get clean data on engagement without contaminating your primary paths. Automate the evaluation: set a logic node that monitors cost-per-trigger and drop-off rates. If the numbers hold, you can scale across other journey nodes. If not, the automated win-back flow kicks in and you pivot without manual firefighting.