PPCChat | Let’s Have Some Real Talk About AI & PPC

Posted September 23rd, 2026 by Amita & filed under PPCChat.

During this week’s PPCChat session, PPC professionals shared their thoughts on AI companies’ responsibility and potential legal liability, how AI could be regulated, the rise of AI agents performing tasks on people’s behalf, and more.

Q1: What are your thoughts on how responsible and/or legally liable AI companies are or should be for their models behavior and actions?

I think you all know how I feel about AI in general and I was pleasantly surprised to hear the US treasury secretary say this! @NeptuneMoon

sooo many thoughts. LOL Absolutely legally liable and need to be held more accountable when irresponsible (which is the majority of time) @ynotweb

They should absolutely be liable for actions. And for stealing training data. @NeptuneMoon

I haven’t thought deeply on this, and I’m open to having my mind changed. For now I think people who use AI should carry the responsibility for what they do with it. I think the AI companies should be liable for unauthorised use of our knowledge. @Pete_Bowen

TW: Gemini is already being sued for fatal decisions people made while using the app @Chriskostecki

Well it’s a lot more than zero! The push towards usage means there is at least some accountability, and this nonsense of ‘we don’t know how it does what it does’ is ridiculous. I do think a person who chooses to use a tool – show me how to 3D print a gun, or show me bomb schematics – shows intent, but the platform is SO eager to please and so likely to go to far lengths to answer questions means the liability should lie with the model creator. @JuliaVyse

I do not buy what I call the AI 2nd amendment argument (and do not buy it with firearms) – technology needs to be weighed by the harm it causes not the good it causes – all the good does not equal just a bit of harm @Chriskostecki

the one who is behind the wheel is responsible, simple. you can have the world’s fastest supercar but if you crash – you are responsible. works the same way for everything else like weapons, etc. @alimehdimukadam

I think we need a massive conversation about consent in the AI era too. For example, I absolutely do not want someone I text with allowing an AI to have full access to our texts. And they can do that without my knowledge and I think that is also wrong. @NeptuneMoon

it is my understanding pretty much all devices are reading texts and offering summaries. @Chriskostecki

I like @Chriskostecki‘s description but I feel like if you extend it to other technologies like electricity or SMS messaging it’s harder to find a clear way to apportion harm. Do we blame the generation company when someone shocks themselves? Do we blame SMS tech when it’s used to detonate a bomb? @Pete_Bowen

The whole notion of ‘intelligence’ is wrong in the first place IMO. human attribute to siri on steroids or calculator for words. I have a lot of thoughts here. @alimehdimukadam

@alimehdimukadam What about a car that catches fire or a gun that misfires due to defect? How about basing decisions on getting erroneous information? @ynotweb

I think it is very disingenuous of AI companies to throw up their hands when there should absolutely be safeguards in place that prevent the systems from providing clearly dangerous or harmful responses – like how to build a bomb or how to hack an electric grid etc. @NeptuneMoon

I mean, they should just call it “fake intelligence”, its more accurate. @ynotweb

The one exception to my platform agnostic rule: not all platforms are as cavalier with consent and forcing AI. If you haven’t checked out the Microsoft Humanist AI Code of Conduct, I think you’ll find it interesting and a useful framework to assess how much AI is being responsible or irresponsible (this applies to Microsoft as well. @navahhopkins

Erroneous information can’t be penalized – the whole internet will shut down @alimehdimukadam

The agent should have a responsibility to report it if it does not want to be an accessory @Chriskostecki

Anything that a mandatory reporter would have to alert authorities to should be an automatic shutdown and reporting in an AI. And yes, the AI companies have LIED about what their models do and how they function. The CEO of Nvidia actually called them out on it a few days ago. @NeptuneMoon

This is personal here in BC after Tumbler Ridge. Of course the ultimate liability is with the shooter because a chatbot can’t hold a gun. But the failure to disclose, the failure of warning signs, the clear mental health issue, the cross-border disclosure conversation is important. @JuliaVyse

I am pushing for a humanist code of conduct irl without much luck. @Chriskostecki

@alimehdimukadam false. Truth in advertising laws are still relevant, as an example. @ynotweb

AI companies need to be semi-liable for their models, in that it’s their fault if they’re wrong, but I hope we don’t use the whole thing to move to a super litigation world on this. @revaminkoff

Defective product laws also apply… just sayin’ @NeptuneMoon

Is google or bing liable for search results? @alimehdimukadam

Why does it have to be about who you can sue? @revaminkoff

@alimehdimukadam yes? at least their investors think so @JuliaVyse

I think they should be responsible, but until the government actually starts enforcing laws on these corporate giants (Alphabet/Google, Meta, Microsoft, OpenAI …) they’ll all keep publishing codes of ethics and then doing whatever they want without consequences. Nobody can stop them, even if you “win” in court it takes 15 years and oops you’re bankrupt before you can enforce any order. @DanThies

No technology is perfect. @revaminkoff

There actually are debates about how responsible search engines or social platforms are for the content on them @alimehdimukadam. And the conclusions here in the US are murky. Basically these companies want it both ways – to not have to be liable but also to not have to really monitor in any serious way what flows through them. A court case has not yet fully decided where that line should be as these companies function now largely as de facto utilities but are not regulated as such. @NeptuneMoon

@alimehdimukadam we’re talking about the CONTENT of the “search results”. If the end website is subjected to laws, then an organization presenting information as if it is “THE” answer then it is also subjected to consumer laws. @ynotweb

@JuliaVyse Nope and the disclaimer and I agree to terms and conditions in the fine print or else all pizzerias would have sued google when it showed how to make good pizza by adding glue, Layered
my point is – is excel responsible if you use it to show a different view statistically or manipulate numbers or if we increased the budget on a campaign 10x is google liable. @alimehdimukadam

this is just a software that speaks . that’s all @alimehdimukadam

@ynotweb There are 1st amendment maneuvers in there because you can’t apply it to a machine without applying it to humans first (https://f7f763a9.aina-article-preview-words.pages.dev) @alimehdimukadam

The first amendment does not apply to interactions between individuals and companies. It is only about the government restricting individuals. @NeptuneMoon

@alimehdimukadam while I agree the user does bear responsibility for their actions, I just don’t see LLMs as equivalent to excel, or a paper and pen. I don’t know if the US court system is the right arena to solve this, but I do see the push to usage and the flagrant fabrications from the LLM creators as big factors. @JuliaVyse

I equate them to cars, the driver is responsible. @alimehdimukadam

So is the manufacturer for performance and disclosure. @JuliaVyse

in excel or on paper, you are carrying out your ideas, whether they are good or bad. with ai, it is carrying out your ideas so you are both culpable imo @Chriskostecki

We’re all in advertising in this chat. We know about the power of persuation. If there really were only lone actors doing their own thing for which they are 100% liable at all times, there would be no such thing as fraud, as ponzi schemes, as involuntary manslaughter, or culpable negligence. As these concepts do exist, there is room to place accountability on a very pushy, persuasive tool maker and the user at the same time. @JuliaVyse

Q2: What, if anything, do you think should be done to regulate AI?

well, for one– NOT allow the AI tech bros to be in charge of the regulation. @ynotweb

Big science fiction fan here. I think that Asimov was onto something with the 3 Laws of Robotics. We need something similar so that AI models have guard rails that actively have them considering if their actions will lead to harm and how to prevent/mitigate harm @robert_brady

Existing applicable laws should be enforced as step one. Step two should be legislation that is NOT drafted by AI lobbyists but crafted with expert input and much public debate. @NeptuneMoon

AI lobbyists and big tech should not be leading the writing of the laws, though I’m not sure who a good educated neutral party on this fast moving pace is, and whoever writes this needs to understand the nuances of the situation. @revaminkoff

I agree with @robert_brady we need to come up with the parameters of what we want to technology to do, and not give those that control it all the say. Once those are defined we can determine how to regulate it. @Chriskostecki

yeah, if we just enforced applicable consumer, compliance, and privacy laws to these companies, we’d be a in MUCH better starting point @ynotweb

I’d like to see some anti trust enforcement too for collusion between these companies. And we could start with academics and ethicists on legislative ideas @NeptuneMoon

Lots of “we” being mentioned. I’m with @ynotweb here. @Ryansappington

I do think people need to still be held responsible for their actions with AI as individuals – the answer can’t just be to blame the companies. @revaminkoff

At this point we know AI is not 100% – everyone should know the info is not 100% accurate – at a certain point, people need to be held accountable for their own failures to confront that fact if they choose to willingly ignore it. @revaminkoff

See I don’t think the average person who uses AI knows how wrong it is or how often. We do cause we are swimming in tech info. The average person? Not so much or at all. People are dying from following AI advice or instructions @NeptuneMoon

This is like giving a toddler a motor vehicle and expecting no one to get run over. @ynotweb

those that use ai tend to trust it, that’s why everyone says to use it for something other than what they do. @Chriskostecki

There will be IAEA type of body for sure. Every single tool we use – from internet, telecom, TV, cars, planes, trains, food – there is a regulatory body in all countries. And just give all data centers a separate grid/electricity – the plug for which can be pulled anytime. ‘AGI’ is just a switch off in a way at its core @alimehdimukadam

The US military averted a near disaster bombing when AI was wrong last week. @NeptuneMoon

“Shall we play a game?” @ynotweb

They just added an AI layer to air traffic control here in the US and it is in UK as well. @NeptuneMoon

@NeptuneMoon right? what could go wrong? glad I don’t live near DC @ynotweb

I still think that AI companies need to disclose all training data that goes into their systems. Right now we’re eating sausage that not only do we not know how it was made, we don’t even know what the ingredients are. @robert_brady

Enforcement is possible in the EU. In the US, we can make all the “rules” we want but we simply do not have the ability to enforce them against anyone but the smallest players. We can and should force them to build their own generating capacity for data centers. @DanThies

Has anyone reviewed their health records when AI is used? Cause my mother-in-law nearly died from an AI hospital error. @NeptuneMoon

maybe it will self destruct and we will all laugh over this crazy time @Chriskostecki

We need to make sure people are still thinking and not just relying on the AI. and that the next generation is still taught to think. @revaminkoff

I found it and it was caught in time, but what if I hadn’t been checking? Most don’t know they need to or even if AI is being used of how. Again, informed consent it not happening. @NeptuneMoon

But honestly the core and crux is – is a speaking software – intelligent? @alimehdimukadam

Well, the AI companies need to stop lying about their models being AGI. They are absolutely not that, nor even close. @NeptuneMoon

Why are we attributing human traits to machines? @alimehdimukadam

It’s only AGI if it turns out that we’re just word guessing machines too @DanThies

I don’t think smart campaigns are smart neither. @Chriskostecki

@alimehdimukadam we don’t have a great metaphor otherwise, and you can’t raise many trillions of dollars by saying ‘wow word guesser’ @JuliaVyse

I tried to get Claude to help me write some ad copy. It refused. Suddenly had developed morals. @williamhboggs

@williamhboggs supplements is a tough industry. @Chriskostecki

I was down this rabbit hole a while back. Different cultures have different meanings and attachments to it along with language and language unfortunately/fortunately is the most powerful (marketers know). For me, Target CPA was AI. Genuinely @alimehdimukadam

Is Target CPA not AI? @revaminkoff

I wish I was joking about the morals thing. I asked it “So you’re sentient?” and it responded “I genuinely don’t know”. Starting to get kinda sketchy lol @williamhboggs

@revaminkoff Machine Learning Algorithms able to process complex information and predict – explaining all of this along with vector embeddings and semantics etc. – AI @alimehdimukadam

@alimehdimukadam But isn’t Target CPA theoretically doing some of that? Taking all the customer and auction data in real time and using it to determine who is most likely to convert at what cost? @revaminkoff

I don’t think it’s fair to say “Target CPA” and “Target ROAS” are AI unless you the human didn’t put in your goals. Max Conversions and Max Conversion Value are closer to “AI” in that you are trusting the bidding algo to work with your inputs and come up with profit. @navahhopkins

The same way we trust an LLM to give the answer based on our input? @alimehdimukadam

@navahhopkins Fair. @revaminkoff

Essentially – and where do we draw the line on “AI’? Are scripts AI? Do they not get the same scrutiny just because a human made the script? @navahhopkins

A core principle thesis of something I am working on is AI = Amplifying Intelligence the caveat – the intelligence is human so using AI – in my worldview is Amplifying Human Intelligence ‘LLM’s being a tool. @alimehdimukadam

We starting losing the plot when machine learning inside a platform like Google Ads started being called AI. It has very much muddied the waters unnecessarily. @NeptuneMoon

Q3: What are your thoughts on AI agents performing tasks on people’s behalf such as making purchases? You can answer this from both your personal perspective and from your perspective as an ads strategy manager.

I already have a wife to do that for me. @Pete_Bowen

I think it is going to not be widely adopted for a while, if ever. And banks and insurers are going to be a big reason for that. Maybe I am wrong and young people will be foolish enough to let it at their bank accounts and credit cards? @NeptuneMoon

People turn over targeting and messaging to the networks and then pay for the privilege, so anything is possible @Chriskostecki

Oh boy. well, from a paid ad perspective, isn’t this already happening for people who allow “auto-apply” recommendations in ad accounts? LOL @ynotweb

So interesting! My CEO asked my perspective on selling to machines not just selling to customers and it’s a very strange moment. I have many thoughts and not all of them are in harmony. The idea of agentic shopping makes sense for Amazon, Instacart, etc, but why would a medium local biz want to structure their data for these different bots when they could invest in personal connections with their customers? And then the big players, why oh why would Meta Google Amazon all collab to make their robots work together for YOU? they won’t. it’s about to be a MESS @JuliaVyse

Well shoot, we are docking, so I have to head out now. have a great week everyone! @ynotweb

I can see merchants getting hit with a load of returns because AI bought the wrong thing and banning clanker purchases @Pete_Bowen

Maybe there has to be some human validation at some point in the transaction, though I am sure there will be an “accept all” option @Chriskostecki

I think it depends – we already do auto-subscribe / auto-purchase on some products. What’s the difference? @revaminkoff

I do think if these become more widely adopted it will drastically change advertising as it will have to be effective in all the traditional ways we are used to but also be very efficient to make the machine choose your brand as the “best” option to purchase. @NeptuneMoon

@NeptuneMoon and how to define ‘best’ @JuliaVyse

@Pete_Bowen Yes on the returns and chargebacks which is why I think it might get slowed from that side. @NeptuneMoon

if you assume the user is conversing with the agent and the agent is acting on its behalf, then yeah it has to be if you want to deal with the logistical issues. @Chriskostecki

Exactly @JuliaVyse and I do not know. But a machine comparing options is not the same as a person comparing options. @NeptuneMoon

I refuse to shop for groceries online, but I am in a shrinking segment @Chriskostecki

I view me using an app for curbside groceries (which I LOVE) not as an agent in this context. That is me making my order myself and the store fulfilling it. An agent would somehow “figure out” what I needed from the grocery store and do the order for me. That is a big difference. One I am the one actively doing the thinking, choosing and approving. The other it is a machine doing all of those steps. @NeptuneMoon

I lean towards AI but this will be such a disaster micro transactions maybe – that’s all. I won’t give an agent more than $5 in a card today. And I don’t think I will be able to give it more than $50 ever
the same way I am for buying a maserati. The driver is responsible as long as you know what you are doing with the tool or like I say Siri on Steroids @alimehdimukadam

Right, but I need to see the shelf and make the selection, I hardly trust my partner to do it @Chriskostecki

@alimehdimukadam I think you are giving non tech way more credit than they deserve in being able to discern about this tech. They only see the hype and not everything we tech oriented folks see. That to me is a big part of the problem developing. @NeptuneMoon

I do see a lot of concierge style supports in the luxury space, and in auto, likely travel, anywhere with a lot of decisions to make before purchase. Then it will become premium to meet someone in person and swing back, but with helper bots across industries. If we live. @JuliaVyse

@JuliaVyse Looks like someone is using Instinct. @NeptuneMoon 100% That’s the crux and it’s marketing unfortunately @alimehdimukadam

And I do think there are technical issues coming for the agent to agent piece. Many businesses struggle mightily at exporting a decent customer list. Having their web sites really prepared for agents to inquire and buy? YIKES> @NeptuneMoon

@alimehdimukadam I think Instinct, Muse, and others like it are very likely the next wave of products. @JuliaVyse

I shared an example of Copilot works in #all-things-ai which I’m genuinely curious how this group feels about it. I personally love one click check out and have been using for years, but I also really like that humans have choice in going to the site to verify vs buying directly on the AI surface @navahhopkins

Agentic shopping has the potential to be abused on a scale we’ve never seen before. So many places for developers to put their thumbs on the scale at the expense of consumers. @robert_brady

I think Muse will take off given the sheer scale of distribution @alimehdimukadam

Maybe they can find all of the traffic lights in the photos then!! @Chriskostecki

As I said, I think banks and credit card companies are going to play the biggest role in how this evolves. They will not just let things run wild and be doing massive chargebacks daily. @NeptuneMoon

@robert_brady the prompt-injection wars will be fought at checkout @JuliaVyse

@alimehdimukadam And I definitely don’t want my agent coming from Meta. Way too many conflicts of interest there. @robert_brady

I find it absolutely WILD that people would want to give Meta even more access after their history with data and ethics. @NeptuneMoon

@NeptuneMoon then again, agentic shopping through paypal via THAT GUY sort of circumvents the banking oversight. @JuliaVyse

This whole slow down the pace of frontier and everything has a very conspiracy theorist angle to it as well. Honestly – all of this is just a war of narrative and marketing @alimehdimukadam

Some argue the whole slow down is so that they can do their IPOs before it all inevitably implodes. @NeptuneMoon

Elon had good timing, coz the more they delay the more valuation drops as open source is adopted. In my opinion – this is the only way they know how to do marketing and be in the headlines. If i tested my F1 car on a highway, I won’t be complaining about how it crashed and caused an accident but the anthropomorphic headlines help so ride the way. Can be 100% genuine – not questioning anyone but we know what sells and gets amplified and how to work the algos. @alimehdimukadam

Q4: Are any of the brands that you work with actively doing anything to try to be attractive to AI agents?

Anyone investing in structured data/fixing their feeds is arguably doing this. @revaminkoff

Yes, but my NDA has entered the chat. @JuliaVyse

I think all the extra fields in feeds are a part of prep for this too. There has never been a time when clean data was more important (though it always has been!) @NeptuneMoon

Yes – we haven’t touched on AI lead gen, they may not shop but schedule and fill out forms is already happening @Chriskostecki

I can share that some of my public sector and health clients are very interested in cross-page actions and form fill ease. Basically how do we surface the good information people need? @JuliaVyse

WebMCP is the new AEO @alimehdimukadam

Q5: How are you feeling about the role that AI is playing in the work you do on a weekly basis? What is good? What is terrible? What is exciting? What is exhausting?

When I use AI a lot I get to the end of the week without feeling like I’ve accomplished something, even though I’ve done a lot. @Pete_Bowen

Unsurprisingly, I am not an AI user. I am very tired of having to read through AI’s analysis of PPC performance or strategy. @NeptuneMoon

I could not do all the work I do without AI to help me, but the AI I use is my choice and I don’t delegate creativity, analysis, or reasoning @navahhopkins

I really like that I have these tools to help me. I really wish they were better at what they did. I really wish there wasn’t so much do-si-do to do when client tools aren’t our tools and vice-versa. All in all, they’re cool-ish, but I would prefer 6 smart interns I can teach and talk to than my two token goblins. @JuliaVyse

had a client abruptly declare his 10 year old crm is dead and sent me a claude login last week, so par for the course. @Chriskostecki

@Chriskostecki there’s no emoji for the face I’m making, but you can certainly picture it. Closed-lip-turned in smile, tired nod, you get it. The GenX resigned nod and sip. @JuliaVyse

I am also constantly astonished at what people just willingly upload into an AI. With apparently no concept of that data now being fair game in the system for usage. @NeptuneMoon

Otherwise I resist it, i do not see the value and focusing any free time to shoring up processes and documentation. maybe at some point it will converge. my employees and resources use it, and i am not against it. i worry for my kids mostly. @Chriskostecki

@NeptuneMoon oh wow, Sophia doesn’t care about the cloud AT ALL @JuliaVyse

Learn to drive – the cars will keep changing not different than a calculator, excel, or computers in general. The only difference is that this calculator speaks in words instead of numbers. @alimehdimukadam

@JuliaVyse yes, as a fellow gen x’r I know this face well. @Chriskostecki

I literally have no choice. Like not mandated or anything, but I have no option besides working to midnight every day. @JuliaVyse

I nearly made my insurance agent and lawyers heads explode when I sent them a bunch of AI liability questions. And my next thought was is no one else thinking about this or asking these questions??? @NeptuneMoon

@NeptuneMoon probably not. I wrote the AI policy for my strata and was VERY pleased to see an AI statement from my accountant. It’s not on everyone’s mind yet. @JuliaVyse

@NeptuneMoon There are a few insurance companies startups you should be investing in coz this is going to be a line item starting next year for sure. @alimehdimukadam

@alimehdimukadam names please, my wealthsimple needs a boost! @JuliaVyse

I read a tweet that the biggest validation of threat is insurance underwriters and what are they pricing it at. in these companies pvt rollups there is something else. @alimehdimukadam

All good, I’ll ask perplexity! @JuliaVyse

Kids attitudes on AI are interesting right now. I have an 8th grader. Some absolutely hate it and use “that’s AI” as a callout for bullshit. Others are dependent on it like an addiction. I am worried that there are more addicts than skeptics… We need the next generation to know how to actually think and write and do basic math, etc. without the help of AI! @NeptuneMoon

AI is great for keeping track of things, especially when you work with dozens of people through dozens of different systems, with multiple Google workspace emails/logins. Basecamp, Slack, Roam, Monday, Teamwork, Clickup and that’s just the first 6 that are in my log from this morning. @DanThies

PPCChat Participants