AI is changing PPC, but how much should marketers trust AI-generated recommendations? In a recent PPCChat discussion hosted by Julie F Bacchini experts shared their experiences, highlighting the importance of human judgment, context, and experience when using AI for PPC.
Q1: Have you seen an uptick in clients, stakeholders or potential clients coming to you with AI generated ideas or strategies? If so, how do you handle it?
I feel like nearly everyone is doing this and it is aggravating. @NeptuneMoon
Definitely getting more Claude outputs handed to me. Some are ideas where the client did some extra legwork to flesh it out with AI. Some are more confrontational since the Claudit said something negative and they want an answer about why we did things the way we did. @robert_brady
Yip. I was the victim of an AI audit ambush a while ago and handled it completely wrong – I tried responding point by point. My current approach is polite interest paired with procrastination. @Pete_Bowen
Just read one yesterday that painted a fairly negative picture because current efforts used more broad match (phrase and exact did so much better in the past) and CPCs had increased, leading to higher cost/lead. Without proper context around the evolution of match types and ever-increasing CPCs, this would have been damning stuff. @robert_brady
Oh yeah, clients are putting their data into Claude and getting feedback on the account. it is the reality we are in and it is not going to go away, so I just say thanks and that I will look at the data. @Ichasse
I have a context and verification setup and communicate it to the client directly which I tested myself again – the broader summary – the machine doesn’t know what your goals and budgets and constraints are – it is recommending what it feels the average person should do by summarizing information. Does it know your backend capacity and or what we decided last month that your CRM is broken and we need to fix that first? or does it know you can’t spend more than 20k per month and why? If it doesn’t then you know the answer. @alimehdimukadam
When they have Claude of whatever platform review performance over the years, it is going to look rough due to the changes that Claude does not take into context. That context can be everything. Like Google raising CPCs and other things. Things that if we put the data into the tool and then put context behind these AI systems would say, okay I get it. @Ichasse
An agency I work with has an AI tool that scans all client accounts and makes suggestions. So far 90% if its feedback has not been accurate and has just made more work for me as I’ve had to defend my stance. It even once suggested I make illegal claims in ad headlines. That said, I do like AI when you give it very specific guardrails. I have a few killer Claude Skills that help me work more efficiently. I think what separates pros from amateurs in the AI era is knowing when to use AI and when not to – as well as when to trust it and when not to. @williamhboggs
We get it somewhat, but more often it’s about strategic inputs than specific levers to pull. @JuliaVyse
Just yesterday, I was given a whole roadmap. Maybe 1 or 2 things that were somewhat valuable, but a lot of the recommendations were to undo things we’d just recently done because claude doesn’t have all of the historical in-platform context. @AnthonyMcDaniel
You can get ahead of this type of stuff by showing that you are also using AI to help identify opportunities. I mean if they are okay with uploading 5 years of data into the tool then obviously they don’t have any privacy concerns, lol. AI can go through 20,000+ line spreadsheets a lot faster than we can and find trends. @Ichasse
One useful thing has come out of this, and that was solidifying the concept of asymmetry in my mind. An AI can produce legit-sounding content far faster than a human can analyse it, and fighting it puts me on the wrong side of the asymmetry. Not a place I want to be. @Pete_Bowen
Yes – AI is designed to always sound confident and that makes this challenging. Even when it is wrong, it sounds like it should be right. Which leads people with less knowledge to question things feeling like they have an understanding they do not. @NeptuneMoon
We do need to let them know once they upload that data unless they have an enterprise account with the AI tool (some of them), your data is not private any longer. This may mean they need to update their T&Cs, etc… This may also mean they are violating some privacy rules with some states, so they really need to tread carefully at the end of the day. AI is a third party, so you are sharing data with a 3rd party in most of the tools (there are acceptions). @Ichasse
@Ichasse And that is IF the AI tools actually keep their promises to not train on data input through an enterprise account @robert_brady
We’ll have a huge run and influx of open source models and any client who wants to use LLMs will have their own setup running on premises with an open source model chinese or not. @alimehdimukadam
If AI has no context they will say why aren’t you using more ETAs, regular shopping campaigns and exact match if they are looking a few years back before Google and the other platforms made a lot of changes to help with stock prices, lol @Ichasse
@alimehdimukadam Yes, Kimi, GLM, Deepseek are big ones right now that are open source. This is going to get bigger as pricing of the other LLMs gets more expensive. @Ichasse
I foresee a huge increase in both token usage and token costs for these tools. If an AI is truly able to replace a junior programmer for example, then the pricing will increase until it’s juuuuust below what that junior programmer would have cost the business. The days of a flat subscription getting you meaningful results will end quickly. @robert_brady
@robert_brady great point. completely impossible that these companies will leave money on the table. @AnthonyMcDaniel
@AnthonyMcDaniel Especially after the hundreds of billions they’re investing in compute. Investors will eventually demand some return on these outlays. @robert_brady
That’s part of the coming crash IMO. The actual cost won’t be supported by big Cos when they can hire juniors more cheaply and get kudos for creating jobs, and the open source model can be in-housed with privacy rules in place. @NeptuneMoon
I’m getting less of this at the moment, and more of “look what this $200/mth tool can do”. It’s everything from full-blown SEO to highly qualified B2B lead generation, all from AI companies less than 12 months old. @TJKropp
The crash will be in valuations of OpenAI and anthropic – not necessarily the infra buildout open source still means you need the infrastructure of gpus and data centers and electricity and everything else. It’s just the brain being offered for free – you still need the body and the body will still need power and energy to sustain and run. Google cloud is the winner here – sleeper giant but all those cloud providers and nvidia are going to win more. @alimehdimukadam
@alimehdimukadam I mostly agree, but data centers are already obsolete in certain cases. The ‘data centre’ is going to be on device in a big way. I think it will be like the dot com crash, but we’ll be left with really good plugins for specific things, and big giant models focused more on big data than ads. @JuliaVyse
Yes, and nuance – biggest of which being how many companies will have the capacity and technical knowhow to do it vs just using the gpts and claude’s and Google’s because that same argument is valid for in-house vs agency and why use Google workspace for Gmail when you can build your own custom smtp server and use free emails etc @alimehdimukadam
Q2: Are clients pulling back in what they want from you or even terminating contracts because they think they can manage things for themselves? If so, how are you handling it? And if not, are you preparing for this in any way?
No, actually clients are demanding a lot more insight from us on how these models work and in managing risk assessment. Part of my role is advising on ChatGPT ads and on using AI tools.@JuliaVyse
I literally changed all my messaging when I redid my site earlier this year to be themed on “human centered PPC” so I am trying to not fight the AI so much as present myself as an alternative or addition for those interested in human expertise and perspective on their accounts. @JuliaVyse
I haven’t had this yet. @Pete_Bowen
I also posted earlier in the Google Ads channel that Google is coming for us with all their AI in the platform: https://blog.google/products/ads-commerce/google-ads-analytics-AI-updates/ @NeptuneMoon

Not yet but more about what more can they do – and it’s usually that Jevon’s paradox playing out but the blockers are no longer what we can do, it’s what they are ready for – legal included and what their policies allow or doesn’t @alimehdimukadam
I’m changing the work I do to be more about the plumbing than the actual ad management. I’m hoping it’ll be more durable. @Pete_Bowen
Anyone confident enough to use an AI instead of a PPC expert isn’t a good fit as a client for me anyway. One of my biggest selling points is experience/expertise from doing this 20 years @robert_brady
Not yet, but it will be coming I am sure. I suspect folks are going to build out tools that use AI and will build out accounts for you and manage them if not already, very soon for a cheaper price than hiring a professional person to do it. @Ichasse
What’s old is new again: does anyone remember Albert AI? I did a first to market test with them ages ago and it went….fine? but nothing special. Reminder that part of this is gold rush thinking and the tide will go back out. just to mix a pile of metaphors. @JuliaVyse
Even big agencies I suspect will create agents to do tasks instead of hiring junior marketing folks. This will hurt in the long run, because we need to train junior folks to one day become senior folks who can work in tandem with AI tools to give you the best results. @Ichasse
@Ichasse I see you, and your thinking is right, but now that we’re seeing some of the ‘results’ we get from ai….those junior roles aren’t going anywhere. They’re just going to get paid way less @JuliaVyse
Q3: Claude is about to include a watermark in everything it outputs so you can tell if things were done in AI for EU compliance. Other AI companies will likely follow suit. Will you check to see if client or stakeholder communication were AI generated? Do you have concerns about your work being checked in this way (whether you use AI or not)?
https://techcrunch.com/2026/08/11/anthropic-says-it-will-watermark-text-generated-by-its-ai-models @NeptuneMoon
Probably, but we like to declare anyway. I like to just straight up tell people when I’ve used an ai and what the sources are. @JuliaVyse
I think this is so interesting. @NeptuneMoon
Honestly, folks will just pull the data from AI and put it in their own document. There are always workarounds to this and it will be done. @Ichasse
I don’t use AI for output, so I am not concerned about my work being checked on this front. @NeptuneMoon
I have concerns about my writing being flagged as AI-produced because AI was trained on what I wrote. @Pete_Bowen
I might be curious if a client was not up front about using AI and being able to check this @NeptuneMoon
It is going to be very interesting to see how this all plays out. @Ichasse
I’ve been accused of being an AI but that was from a youth with the attention-span of a gnat. They couldn’t fathom that a human could produce 600 words in a reddit answer. @Pete_Bowen
I like the idea of a watermark. Europe’s compliance infrastructure has some benefits because the penalties are steep enough to force these companies to do something. Interested to see if it catches on @robert_brady
In the EDU space I can definitely see proficient human writers being accused of being AI. Technical writers in particular. @JuliaVyse
I actually kept proof of work (the printouts of my early drafts and my red-lining) when I wrote my book last summer just in case. @Pete_Bowen
I hope my writing/outputs convey me enough that clients will recognize it was me who wrote it. @robert_brady
I hope to be at least semi retired before this all fully fully manifests @NeptuneMoon
Maybe if I throw in enough Fboms and hockey stick emojis it will feel authentic? @JuliaVyse
For me it’s all the parenthetical notes (I’ve always got a caveat or aside that has to be included in a thought) @robert_brady
Google already has synthid for images and videos and EU rules have gone in effect so that watermarking will be I think synthid which is being made a standard by Google fun fact – Google also owns approx 10% in anthropic – not sure of exact %. @alimehdimukadam
They’re just handing each other a trillion dollar IOU and pretending it’s on all balance sheets at the same time. Mark my words: these five companies, and the bros who run them, are about to run face first into annoyed investors. @JuliaVyse
Q4: Are you doing anything differently in your business or career development to try to combat the wave of people thinking they can manage their PPC strategy themselves with the help of AI?
Honestly, part of the strategy involves waiting out people trying to do it themselves with AI and finding it is not as great as working with a professional. I am sure there will be plenty of advertisers that go this route and stay with it and they are not the ones for me anyway. @NeptuneMoon
I don’t actively try to fight this 1:1. I do like to publish my thoughts and host info sessions. I can’t convince someone not to hire a robot, but I can be there when they learn the hard way that it doesn’t work, isn’t compliant, and actively creates risk. @JuliaVyse
I was saying to an industry friend recently that we are living in a kind of perfect storm of people wanting to DIY with AI with platforms pushing their AI with the messaging of management being easier than ever plus a shaky world economy. That combo is brutal. @NeptuneMoon
AI if used and monitored correctly could be a great help to the kind of company that needs it. And we have ai-tools inside my org that work really well for us. none of them replace a person, and they aren’t sustainable. @JuliaVyse
Part of me likes your stance @NeptuneMoon Let them FAFO and then they’ll need someone to help clean up the mess. @robert_brady
Cory Doctorow has a good way of putting the expectations around AI: It’s like breeding faster and faster horses and hoping one of them will give birth to a locomotive. @JuliaVyse
I have been through rough cycles before @robert_brady and one thing I know for sure is you cannot force people to move through something like trying AI to DIY. You do have to wait for the FO stage. @NeptuneMoon
Remember back when AdWords still used their “Set up your account in 5 minutes” language trying to make it feel so easy for the DIY and SMB folks? There seems to be some parallels here @robert_brady
I absolutely remember that. And today’s messaging is that on steriods. @NeptuneMoon
I’ve been repositioning what I do to be less in the account and more about the overall picture: making sure that the data is wired up properly, making sure that the business is functional. @Pete_Bowen
Yes, it’s as easy as Meta. @JuliaVyse
@robert_brady Weren’t they called smart campaigns or something. You basically just put in your credit card number and your website and google spent your money.@Pete_Bowen
I think we’ve also been trained to think we can DIY everything because of how useful YouTube is in that regard. Except YouTube has actual people with some kind of experience in the thing they’re teaching.
AI just sees some of that content and tries to repeat it without knowing if the output is actually accurate or a frankenstein of different sources & projects. @AnthonyMcDaniel
The word guessing machine is good at word. It not good at think. @JuliaVyse
I embrace AI to an extent. I love it with proper guardrails in place. I don’t let it run my ads, but it has been helpful for a lot of side tasks like scripting listeners for tracking in different platforms. But I agree with the above points from others. I don’t think AI should be left alone to run loose with accounts and anyone who tries this is bound to fail (with AI in its current state. Who knows what it might be able to do down the road). So much advertising has nuance that needs a human touch. But I can also see a case in the future where AI is given a book of guardrails and really can run an account. But I really do mean A BOOK. @williamhboggs
And becuase we evolved our thinking around language, it feels like it can think like we do. @JuliaVyse
I will add that I like Wil Reynolds post on LI about sitting with interns/junior people who are using AI a lot and learning from them. There are two sides to this and benefits on both sides @robert_brady
@williamhboggs yes! a local model with a LOT of training data. Will is very good at think! @JuliaVyse
Let’s also not forget that platforms would LOVE for the dollars currently going to agencies and consultants to instead be spent in the platforms. @NeptuneMoon
On the original question, my main “act of resistance” is essentially to go through these recommendations and point out the areas it’s falling short, or misunderstanding PPC strategy, or missing context entirely.
I’m not sure it’s doing much, unfortunately, since I keep being advised based on LLM output. @AnthonyMcDaniel
Agreed. Reduce the cost of complements to zero. @Pete_Bowen
Marketing is at its core a human function. There may be patterns, but human beings are both emotional and rational in varied proportions at various times. Human beings are still best at interpreting the state of another human being and responding accordingly @robert_brady
The best AI production in our industry is going to come from smart marketers working with AI tools to help analyze data quickly. AI still does not understand the customer the way a seasoned marketer will. It won’t lead user study groups, usability sessions, etc… like we can and do. @Ichasse
@Ichasse I don’t even think that’s the use case yet. AI is for rebalancing wages and prices 100/per min to extract profit. The nuances of behavioural data are really only possible with live mass surveillance. Without TikTok views or real-time feedback, we will always outthink marketing machines. @JuliaVyse
I am actually working on launching a tool that does llm simulations and identifies the errors offline and helps others spot them before they do it live like a driving school but for working with llms @alimehdimukadam
To the “in your career” part of the question. I think AI is pushing us to need to build our own authority and trust. AI can be a useful tool but is our competitor now whether we like it or not, so we have to prove why we as humans are more valuable in your account than the robots @williamhboggs
Q5: The PPC industry has always been a place of regular change, but the changes are coming more and more rapidly lately. How are you handling it?
Barely @robert_brady
I just had a chat this morning about this. The new normal is ‘this thing is changing next month’ or ‘today! we’ve updated your favourite tool, suck it!’ @JuliaVyse
This year for the first time I feel like I could just pack it in and retire if my wife wins the lottery. (Assuming she stays my wife after that) @Pete_Bowen
I just want to validate for everyone that my statement in the question is true and things are changing faster than ever. So if you feel some whiplash or exhaustion over that, it’s ok! @NeptuneMoon
@Pete_Bowen if my husband won the lottery I would instantly quit my job to start managing the money. He can play jazz and buy saxophones on his allowance. I’m handling it with comms and plans. I try to keep my news up, my team informed, and client conversations topical. It’s a lot, but it’s also a lot of opportunity to communicate and connect. @JuliaVyse
@JuliaVyse sounds like you’re the perfect wife. I imagine the formula for how many saxophones a man needs is that same as for surfboards. Need = N+1 where N = current number owned. @Pete_Bowen
I am also trying to have more conversations about what is most important to the client’s business and how can we work to make sure they are getting that as effectively as possible as platforms change the rules and backend monthly. @NeptuneMoon
We’ve moved a lot and the current checklist starts with: fill up a car with saxophones (in this case a bass clarinet fits) then put boxes and piano and so forth in the truck. @NeptuneMoon exactly. Some clients need a lot of info, some just want to know you’re on top of it. Ultimately the team needs to know. @JuliaVyse
First principles and back to basics – tools and tactics will keep changing as they always have marketing first principles of what makes people buy and why and how to promote and everything else will always stay the same. the only other big change is advertising to agents – that needs a separate muscle to build @alimehdimukadam
I am tired of explaining how “well yes, that has been working that way for quite a while but now it won’t” @NeptuneMoon
Advertising to agents. How dystopian is that? And so fraught with potential issues @robert_brady
Being part of this community helps a lot! Same with being fairly active on LinkedIn and following a lot of different PPC folks. Additionally, actually participating in dialogue here and on LinkedIn has been instrumental for me. It forces me to actually put my thoughts/understanding into written word, which helps me to actually digest and understand it. That way I know I can relay info to a client as well. It also helps challenge my thinking, and putting my own thoughts out there and asking for critique has been really helpful in rounding out my understanding of a lot of things, including the Aug 17 updates. @AnthonyMcDaniel
If it ever goes to mostly agents, I am out @NeptuneMoon
TIME magazine has already started doing that @alimehdimukadam
I feel like @alimehdimukadam nailed it. @williamhboggs
And perplexity apparently already blocked them saying it’s misleading @alimehdimukadam
I also think if it does go mostly agents, platforms will use that as their just “give us a few parameters, a URL and your credit card” on us. @NeptuneMoon
Stay rooted in the basics like psychology. Build your marketing around that and all the changes don’t mean as much @williamhboggs
I’ve been trying to crack this and deduce it from first principles – feels like I did but will share it with you all to let me know if I became a victim of AI psychosis or not. the figures are estimated costs I would have paid if I had used official API’s – right now it’s leveraging the delta between subscription limits. @alimehdimukadam

It is going to be a bumpy road, just keep your head up and keep learning. Some companies will use these tools as gospel and others will understand your value. Your customers will come from the second bucket. Don’t fret about the folks in the first one. If you do your job, just hope the competitors are part of the first bucket. @Ichasse
PPCChat Participants
- Julie F Bacchini @NeptuneMoon
- Julia Vyse @JuliaVyse
- Peter Bowen @Pete_Bowen
- Lawrence Chasse @Ichasse
- Anthony McDaniel @AnthonyMcDaniel
- Boggs @williamhboggs
- Ali Mehdi Mukadam @alimehdimukadam
- Lawrence Chasse @Ichasse
- Boggs @williamhboggs
- Anthony McDaniel @AnthonyMcDaniel
- Robert Brady @robert_brady
- TJ Kropp @TJKropp
Related Links



Stop the wasted ad spend. Get more conversions from the same ad budget.
Our customers save over $16 Million per year on Google and Amazon Ads.