{"id":21743,"date":"2023-04-19T17:33:02","date_gmt":"2023-04-19T12:03:02","guid":{"rendered":"https:\/\/www.karooya.com\/blog\/?p=21743"},"modified":"2023-04-19T17:33:03","modified_gmt":"2023-04-19T12:03:03","slug":"ppcchat-ai-helpful-or-headache-in-ppc","status":"publish","type":"post","link":"https:\/\/www.karooya.com\/blog\/ppcchat-ai-helpful-or-headache-in-ppc\/","title":{"rendered":"PPCChat | AI – Helpful or Headache in PPC? | 18th April 2023"},"content":{"rendered":"\n

Greetings Readers! AI and PPC are two areas that are becoming increasingly intertwined. During this week’s discussion, host Julie F Bacchini<\/a> sought PPCers views on the aspects of AI that are helpful for PPC, what is on their AI wish list for PPC and more.<\/p>\n\n\n\n\n\n\n\n

<\/div>\n\n\n\n

Q1: How do you define AI when it comes to PPC?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

I don’t think we have true AI yet in PPC. I believe we have machine learning (ML) in lots of places like bidding, keyword matching, asset creation, etc. and many conflate that with AI. @robert_brady<\/a><\/p>\n\n\n\n

It’s not actually AI at this point, more like strong machine learning. Asset creation\/optimization, search\/interest matching. Basically, if you’re ecom, it’s a very powerful work back from bottom of funnel to top. Cool, but not what Asimov was thinking of. @JuliaVyse<\/a><\/p>\n\n\n\n

Echoing @robert_brady<\/a> – A lot of things are being called AI right now & they are rapidly getting integrated into platforms and tools. I don’t think the level of the tech is actually all that sophisticated at this time either, despite all the hype stating otherwise! @NeptuneMoon<\/a><\/p>\n\n\n\n

I agree with this. There is probably some AI driving the ML we are using but we aren’t using AI yet IMO. @TheMarketingAnu<\/a><\/p>\n\n\n\n

Machine learning is now ubiquitous throughout Google Ads and Analytics. LLMs generating possible ad copy, query matching, bidding, and lately even the whole cookieless browser tracking models at play. @gilgildner<\/a><\/p>\n\n\n\n

For starters, I think of AI as different from machine learning. ML is used by the platform algorithms to help serve ads better, whereas AI can be used as a complementary tool to make our jobs easier (like drafting ad copy or automating a manual process) @adclarke10<\/a><\/p>\n\n\n\n

Agree with what others have said. We don’t really have AI in marketing yet. We just have machine learning, which is very different. @lchasse<\/a><\/p>\n\n\n\n

As of right now, a fun buzzword to get everyone excited about something that is probably going to cause you more of a headache than anything else. @JonKagan<\/a><\/p>\n\n\n\n

The distinction b\/w AI & machine learning is important. AI = Autonomous, human-mimicking decision-making ML = Performing specific tasks w\/ accurate results based on inputs (learning data) and pattern recognition To date, Google Ads has not (fully) rolled out AI. @teabeeshell<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q2: What aspects of AI do you think are helpful for PPC?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

I think the most helpful AI capabilities at the moment are in idea generation and iteration, for keywords or ad copy particularly. It’s interesting to play around with for competitors too to see how content differs. @NeptuneMoon<\/a><\/p>\n\n\n\n

It could be great for new search\/interest\/behaviour mapping. And of course, the forever elusive dream: cross-channel attribution. A gal can dream. @JuliaVyse<\/a><\/p>\n\n\n\n

While I’m daydreaming…Imagine mapping out a new feature based on reviews and correlating it with analytics behaviours? @JuliaVyse<\/a><\/p>\n\n\n\n

I really like how the generative language models can shift their voice with a simple prompt. That can give you more breadth of ideas and tones. @robert_brady<\/a><\/p>\n\n\n\n

As digital marketers, we\u2019ve shot ourselves in the foot over the past couple of decades by over-emphasis in trackability. I\u2019m pretty sure we\u2019d be served well by taking a more heuristic approach to data & there\u2019s a possibility that AI could help us make better decisions. @gilgildner<\/a><\/p>\n\n\n\n

I think is where it can get exciting. Finding new markets, ideas, and anything we have not thought of. Being able to look at data and find trends or opportunities we may miss. Ideas for everything from copy, video, imagery, etc. @lchasse<\/a><\/p>\n\n\n\n

I enjoy how #ChatGPT<\/a> supports me to write Ads & extensions when I\u2019m trying to word things differently, i get some great buzz words & sometimes flips a sentence around which really helps. I definitely wouldn\u2019t say it can write it (yet) @_Mizzle_<\/a><\/p>\n\n\n\n

When used for suggestions for ad copy\/keywords and certain bid strategies. @JonKagan<\/a><\/p>\n\n\n\n

I think that generative AI tools will continue to be the most useful for PPC, especially for time-consuming tasks like ad copy, keyword\/audience research & competitor analysis. Other helpful use cases may be spreadsheet\/data clean-up & automating internal processes. @adclarke10<\/a><\/p>\n\n\n\n

So far, I’ve leveraged ChatGPT for the following: – KW discovery – KW grouping into themes – Ad copy variations (directionally) – SQR relevance ranking – GMC title + description augmenting – Custom audience segment (steering) – Landing page content ideation. @teabeeshell<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q3: What aspects of AI do you think are a headache or difficult for PPC?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

The way it’s applied! I’m not interested in a ‘take our word for it’ approach. Show me the stuff, be on the team with me. Deploy this powerful tool in a cooler way. be cool I guess, is the ask. @JuliaVyse<\/a><\/p>\n\n\n\n

Honestly, whoever programs the AI is going to have an influence on what\/how it does what it does. This is both good and bad. Companies have biases and this may sway what results are presented to searchers vs. what is best for the searcher. @lchasse<\/a><\/p>\n\n\n\n

Does hearing about it breathlessly reported ad nauseum count? More seriously, hearing about it being deployed after the fact is a pain. Like Google is doing right now with all their “we’re totally doing AI too!” announcements every other day. @NeptuneMoon<\/a><\/p>\n\n\n\n

The biggest difficulty will be managing expectations. People believe AI is a magic wand. It can do anything and do it better than people. The reality is much more mundane and that leads to disappointment. @robert_brady<\/a><\/p>\n\n\n\n

The misconception that 1. AI can replace humans, unassisted, and 2. That AI isn’t just a giant headache. @JonKagan<\/a><\/p>\n\n\n\n

1) Biases in the data that the AI is trained with 2) Plagiarism & copyright issues, especially for AI-generated creative 3) Combatting misinformation & false claims 4) Blander, less interesting content. @adclarke10<\/a><\/p>\n\n\n\n

A common theme is “Is this going to take my job!?” The short answer is…not yet. So far, these tools save time & mental bandwidth, if understood. The headache is learning how to get them to work well for you. That’s a unique skill set (+ learning curve) for all. @teabeeshell<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q4: What do you think is most exciting and\/or potentially the most useful aspect of AI for PPC?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

By far, LLMs. We bought a premium seat of ChatGPT for the agency and it\u2019s been really useful. And even though I was very disappointed in Bard\u2019s release, it turns out Bard was intentionally pretty limited. I have hopes that they will release a newer version soon. @gilgildner<\/a><\/p>\n\n\n\n

If the platforms can get AI (or ML) to marry up interests and queries and produce high-converting ad copy, I’d be quite excited about that. But – it would also do it for anyone who asks, so the danger of all ads being the same would be real. @NeptuneMoon<\/a><\/p>\n\n\n\n

Automated testing, automated dataviz in reporting (MAJOR), potential for automated landing page creation, audience curation, and you know, a partridge in a pear tree. @JuliaVyse<\/a><\/p>\n\n\n\n

Kinda mentioned this in A2 as well, but I think that generative AI tools like ChatGPT will continue to get more powerful & produce more useful outputs. This will help speed up more tedious tasks so that we can focus more on strategy, analyses, etc. @adclarke10<\/a><\/p>\n\n\n\n

One that can pull useful insights after watching my most frequent reports. @JonKagan<\/a><\/p>\n\n\n\n

I\u2019m a freelancer so I rarely have people to bounce ideas off, I use #ChatGPT<\/a> to do that, & obviously I\u2019m loving the ad idea support from it too\u2026 overall a great little helper (doesn\u2019t beat a second PPC brain at times though. @_Mizzle_<\/a><\/p>\n\n\n\n

For me, the prospect of automating mundane or time-consuming tasks is quite exciting. twitter.com\/teabeeshell\/st\u2026<\/a> This holds the potential for better, deeper thinking, more time in my day, or shorter working days. All stellar mental health boosts! @teabeeshell<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q5: What do you think is the most frustrating, troublesome or negative aspect of AI for PPC?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

Ultimately ML uses data that\u2019s been generated by real people. Quality output requires quality input. So the whole world can\u2019t totally shift to AI, still needs to be some human creativity. @gilgildner<\/a><\/p>\n\n\n\n

It’s a tie: The comms. The chicken little. Then it will take over. The PMax. The trust the machine. The panic. The immediate trust. The bot battle. The anti-trust. The MESS. @JuliaVyse<\/a><\/p>\n\n\n\n

I am concerned about too much homogeneousness if everyone is using AI for keyword, ad copy and landing page ideas. Although it also creates space for an advantage for those who don’t just use the AI for all that! Training data bias is a real issue too. @NeptuneMoon<\/a><\/p>\n\n\n\n

And then there’s this… It has the potential to really flatten research. Right now you have to either use a tool of some kind or manually compile your research. If AI is doing it all for you, everyone will have the same info. Again, room to be better than AI! @NeptuneMoon<\/a><\/p>\n\n\n\n

This isn’t necessarily specific to PPC, but I tend to agree with the recent letters\/statements about slowing down AI production in general. This tech is clearly very powerful, but can also be extremely harmful if development moves too quickly & we aren’t careful. @adclarke10<\/a><\/p>\n\n\n\n

There is an inherent risk of human operators taking (lazy) shortcuts. The tools are no substitute for great strategy, quality copywriting, or data analysis insights. (THAT’S actually AI.) I’d hate to see ad\/site content quality degrade over time from tool abuse. @teabeeshell<\/a><\/p>\n\n\n\n

I don’t think any of us have time today to go down that list. @JonKagan<\/a><\/p>\n\n\n\n

I think for me – it’s the whole worry that it will take away their jobs. AI still needs a lot of guidance and it only as good as the people creating it. Having said that – I don’t want Google forcing us to use it before it’s ready! @TheMarketingAnu<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q6: What is on your AI wish list for PPC?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

For now, testing, dataviz, and hi volume ad creation IF it can understand specific parameters. It’s a tool ultimately. Probably more on my list as things develop, but that’s on the short term for me. We can decide if\/how to deploy. @JuliaVyse<\/a><\/p>\n\n\n\n

Oh, you know what? FIND AND MANAGE MISINFORMATION. That would actually be great thanks. @JuliaVyse<\/a><\/p>\n\n\n\n

Where to start… I would love a report creator where I told it what I wanted and it put all the data together without me having to select EVERY DATA POINT. AI alerts based on what I tell the platform is important to me. Better recommendations. @NeptuneMoon<\/a><\/p>\n\n\n\n

I don’t hate the idea of ML\/simple AI media planning. Hey planner pal: emphasize local media in my next buy, or tell me what mood people are in on the various platforms I use. @JuliaVyse<\/a><\/p>\n\n\n\n

More advanced semantic understanding of words. How about some basics like not showing ads in untargeted geographies? Understanding query words that conflict w geo-targeting settings & not showing the ad. No more disapprovals for single words cause “trademark”. @NeptuneMoon<\/a><\/p>\n\n\n\n

Suggesting possible negative keywords based on the ones already added to the campaign or account. Highlighting suggested placement exclusions based on previously excluded placements. Ad copy suggestions based on the ads from advertisers in auction insights. @NeptuneMoon<\/a><\/p>\n\n\n\n

To stop calling it AI and just call it automation. Also, a more acceptable way of making changes with required sign-off first. @JonKagan<\/a><\/p>\n\n\n\n

I just can’t stop now that I’ve started! How about a setting to auto-compare against a time frame and highlighting what’s changed and a why from the AI? Like “CPCs are up for this term because advertiser X is spending a boatload on that term now.” @NeptuneMoon<\/a><\/p>\n\n\n\n

There is plenty of room for machine learning to address poor results in search query data. Ex) If my “learning data” is negative KW lists, learn from that, address pluralization, infer & avoid similar queries. Advertisers provide powerful signals to Google, too. @teabeeshell<\/a><\/p>\n\n\n\n

I’d love a way to faster advise my clients. I tend to get into a VERY FUN situation (every dang time) where they want advice on the next campaign before this one is completed. a quick interpolation would be super helpful. @JuliaVyse<\/a><\/p>\n\n\n\n

Surface the individual assets that perform best. Either suggest similar assets (copy) or prompt advertisers to create them (visuals). Currently, the only search text ad asset-level “performance” data is: type, position pinning, impressions, & source. @teabeeshell<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

PPCChat Participants<\/strong><\/p>\n\n\n\n