Happy Sunday, Everyone! (& Happy Fathers’ Day to all the dads out there)!
I spent the last week in Boston for the first US-based, in-person SMX event since SMX West (way back in February 2020), and it did not disappoint. If you have the opportunity to attend an in-person event, I can’t recommend it enough – the energy, the networking, the conversations, the learnings are 10x greater vs. virtual events. The Third Door Media + Search Engine Land teams deserve a ton of credit for pulling together an incredible, well-organized event filled with exceptional speakers and ample opportunities for networking and connection.
The big question I had going into this event was where are we (collectively) in the AI hype cycle - and how close is the valley of disillusionment? After three days of talking to over 100 agency and in-house marketers, brand owners, founders, investors and mar-tech executives, I’m quite confident we’re getting close.
There’s a chasm between the sky-high AI hype (and the highly curated AI demos) that everyone is seeing, and the on-the-ground reality of how these AI systems operate (or don’t) on a day-to-day basis. Put another way: what we see in fancy product announcement videos and what we get from the actual tools don’t align. On one hand, there are executives who believe AI will replace entire marketing, customer success, development and customer service departments (that’s not likely to happen any time soon); on the other, there are front-line employees working with these tools who can’t even get them to produce on-brand copy or functional code with any degree of consistency.
That’s not tenable. Something has to give.
There’s a saying that history doesn’t repeat itself, but it does rhyme – and while tech luminaries like Bill Gates have compared GenAI to the Graphical User Interface (GUI), the more I think about it, the more I think the relevant comparison is the emergence of the internet. Then, as now, hype was writing checks that the tech could not cash. There was a mad dash to make and do everything online – but most of it didn’t work at first (leading to the dot-com bubble bursting). It took another 20(ish) years and 3-5 massive breakthroughs (mobile, 5G, ML/Deep Neural Nets, GenAI), for us to get to a point where the big ideas from the dot-com era were both technically feasible and economically viable.
Seismic shifts in tech happen – but they do so on longer timescales than most tech visionaries want to acknowledge. GenAI is a seismic shift. It’s (potentially) the most important technological race to this point in the history of humanity. But it’s still early.
This all brings me back to the hype/reality chasm and the valley of disillusionment – because it’s clear that we’re rapidly approaching it. The hype (“AI agents that do everything for you!” and “Replace your CMO with AI”) and the reality (ChatGPT can usually generate images with mostly correct spelling and the right number of fingers and toes) are miles apart. Virtually everyone I talked to is actively trying to integrate AI into their workflows, but frustrated that it isn’t as easy, as foolproof or as seamless as they expected.
If that’s how you’re feeling, know that you’re not alone, and that (as I shared in my SMX keynote), the friction is your opportunity. When things just work for anyone at all times, the alpha created by those things goes to zero. If everyone can do something with virtually no effort, there’s next-to-no incremental value in doing it. When things are weird, funky, counterintuitive and unpredictable, most people get frustrated, give up and revert to doing things the way they’ve always done them. Change is challenging under ideal conditions, and what we have with GenAI is far from ideal.
That leads me to point #2: the marketers, brands and agencies that win over the next year aren’t going to be the ones hellbent on building AI systems that can do what the hype videos suggest; to the contrary, they’re going to be the ones that accept these tools as they are, then find ways to use them as force multipliers for their existing teams – automating monotonous and rote daily tasks, then re-investing the time saved into doing more strategic, more impactful work.
Over the past 10 to 15 years, marketing has become increasingly obsessed with lever pulling: tweaking bids, building workflows, running tests, launching automations, and optimizing every toggle and dropdown menu. We’ve spent so much time focusing on what buttons to push that we’ve lost sight of why we’re pushing them in the first place.
Somewhere along the way, many practitioners - whether in digital, search, social, programmatic, or email - have forgotten that we are marketers first and technicians second. The real, uniquely human work that must be done is to craft stories, strategies and experiences that resonate with your target audience and differentiate your brand from everything else out there. Everything else is just details (important details, to be sure – but details that can be delegated to machines).
With that in mind, I want to share 7 use cases for GenAI that we’re using right now to free up time for our team while adding value for our clients.
Let’s dive in.
1. Better Competitive Intelligence
For most brands, competitor research is done once (if at all), then never again. The prevailing sentiment among most marketers (even the senior-level ones) is that competitive research is a dull, boring, painful exercise that they can never find the time to do.
They’re right. There are few things more mind-numbing than clicking through a half-dozen (or more) competitor sites each week, trying to identify the changes made from the previous week/month/quarter.
So, instead of manually crawling competitor websites, pricing pages, or third-party reviews, let AI handle the heavy lifting. Start by prompting ChatGPT or Gemini with your brand context, a list of known competitor domains, and a few qualifiers (location, industry, offer structure).
The model will return an augmented list of competitors (I’m continually amazed that Gemini can find multiple competitors we had not heard of or alternatives we did not consider). Once you’re satisfied with the list, it’s time to move on to step #2.
From here, prompt it to provide a breakdown of current offers, pricing models, positioning, messaging/value props, customer sentiment (extracted from reviews) and social proof (i.e. press coverage, influencers, etc) for each identified competitor/alternative. Have it populate these into a table, with one tab per brand. You may have to fiddle with the prompt to get exactly what you want (see above: friction is your friend), but once you’re happy, head over to Zapier (aside: if you don’t have Zapier or something similar, get it).
Create a workflow like the one below (they’ll get more complex), using your prompt, for each competitor (slightly annoying - but it’s better than getting a bunch of data in a bunch of sheets). Once you have this in a good place, you can certainly upgrade to a better solution – but in a pinch, this will do:
Schedule it to run every week (or month) – and voila! You’ll have competitor insights populated in your sheet every (week/month). While this data is helpful, it isn’t as actionable as it could be – so the final step is to have Gemini/ChatGPT conduct a regular trend analysis of the data for each competitor and flag any anomalies (such as new offers, new messaging, updated value props, etc.) or patterns (competitor B always launches new offers on the 1st of each month).
Armed with this insight, you can spend more of your time determining the right course of action to beat your competition – and less of it figuring out what they’re doing.
For most teams, this is an easy win – either you win back 10+ hours a month that are being spent doing this rote task, or you get actual data that you did not have before at virtually no incremental cost.
2. Genuine Audience Insight + Ad Testing
As with competitor intelligence, most brands struggle to stay up-to-date on their audience. Surveys and focus groups are done when time permits (which is never), and the data gathered from them is rarely (if ever) used to its fullest potential. The reality among most brands I speak with is that their marketing is tailored to audience personas created years ago and updated maybe once since. Effectively - most brands are flying blind when it comes to their audience.
Once again, this is an area where AI can shine.
Start by exporting your lead or customer file, along with relevant other information (source/medium/campaign, last purchase date, first purchase data, total purchase value, total number of purchases, first product purchased, any tags/labels, any other data you can think of for eComm; for leads, focus on the source, date, salesperson, product/service, qualification factors, etc.). Upload it to your platform of choice (I think Gemini is best for this, but you do you) and prompt it to segment the audience. It will likely give you several options (K means, RFM, DBSCAN), of which you can select one. If you’re not sure, K-means and RFM are the easiest. Run the segmentation and export the results.
Upload each segment’s data (email, phones, names) to SparkToro (yes, you can do that - check out Rand’s video here), and it’ll spit out detailed audience information – demographics, psychographics, trusted sites, popular podcasts, frequented social networks and YouTube channels, most followed influencers, you name it. Export that data and head back to Gemini/ChatGPT.
Prompt the model to develop a detailed, multi-dimensional audience persona based on both your segment AND the SparkToro data you’ve uploaded. This should include motivators, challenges, psychographics, behavioral patterns, pain points, interests, sources of information/influence and likely emotional drivers. You’ll likely need to push/prod the machine to achieve the level of specificity and detail you want, but I’ve been able to get something remarkable in about 10-15 minutes.
Finally - upload that entire persona back to Gemini (custom Gem) or ChatGPT (custom GPT) to create a virtual embodiment of that audience segment.
Rinse & repeat for each of your audience segments.
Now, each time you create content, make a new ad or develop a new lander, call one or more of the custom GPTs or Gems (use the @) to evaluate it. See which one(s) resonate with it most – if you’ve created a piece of content targeted at the “executive” persona, but it resonates most with the “do-er”, you likely have an issue. Ask the “executive” what was missing/wrong with the piece, revise it, and try again.
The applications of this are limitless – you can pre-test ad scripts or ad copy, refine content, validate offers/angles, test messaging, even ask which of your competitor’s offers (from #1 above) is most appealing, and how could you craft something superior.
3. Search Term Management That Stops Budget Bleeds
If the first two are more strategic, this one is far more tactical.
Microsoft’s / Google’s match types are a mess. Phrase match is (essentially) old-school broad match. Exact match is anything but exact. Broad match pulls in intent-adjacent queries you’d never bid on given the choice. And while there are some compelling reasons to err on the side of inclusion (all things being equal, I’d rather show ads to the right person who typed the wrong words, vs the wrong person who typed the right words), the fact remains that there are many verticals where specificity matters and where certain words are disqualifying.
The most popular example of this is legal services – in the last week alone, I’ve seen Google match the keyword “birth injury lawyer” to an actual search for “divorce lawyer.” These are two very different areas of law that have next-to-nothing to do with one another; there are virtually no legitimate firms that even offer both. While this is a specific example, it’s hardly a unique one – we see variations on this theme every day in ad accounts. Non-branded keywords (i.e. “best roofer in Maryland”) will be matched to searches for specific companies; keywords for specific products (i.e. term life insurance for 30 year old) will be matched to searches for completely different things (i.e. whole life insurance or annuities). The list goes on and on. Managing search accounts is a never-ending game of whack-a-mole with random, irrelevant queries.
So, instead of wading through STRs line by line, export both your positive KW list (use campaign - ad group - keyword as your dimensions) and your search terms report, upload both to Gemini/ChatGPT, and have it rate each search term on a scale of 0 to 10, with 0 being highly irrelevant to the matched KW/the theme of the ad group, and 10 being an identical match to a KW in the account.
Anything rated 0-4 is a prime candidate for exclusion; if this is a long list, have it generate n-grams of the queries so you can minimize the number of negatives required to exclude this list. All search terms rated 5 - 7 should be reviewed – some of these might be good candidates for their own ad group or might reflect new opportunities; others might just be irrelevant but not egregiously so. Those that are 8 - 10 are probably ok to leave alone.
This is especially critical if you’re running Performance Max. With less control at the surface level, having AI triage what’s actually happening in the background can save tens of thousands a month.
Use this monthly. Or better yet, automate the whole thing.
4. Landing Page Creation in a Fraction of the Time
I firmly believe that the single-most-impactful lever most brands have to improve the performance of their paid media campaigns is the lander. The simple reality is that 90%+ of brands do not do enough landing page testing – and that’s usually because the brand/creative team has a certain idea of what a lander should look like, and is not open to deviating from that pre-conceived notion.
The problem is that (in the vast majority of cases), the customer isn’t a marketer. There are far too many brand marketers telling their audience what LP experience they should want instead of creating the LP that gives their prospects what they actually need to take the next step.
The solution is twofold: (1) leverage that competitive research + custom Gems/GPTs to identify what lander types are most likely to resonate with a given audience segment, then (2) use AI to compress the strategy-to-execution cycle.
Start by prompting Gemini/ChatGPT to create a landing page wireframe based on your brand, product/service, audience, and offer. Ask for copy blocks, testimonials, benefit callouts, and headline variations. Once you have this, call your custom GPT/Gems and ask for feedback – what content is missing? Are the testimonials resonating? Is the story compelling? You’re effectively using the custom GPTs/Gems as an on-demand focus group to quickly get your LP into a good place.
Once it's there, use Midjourney to visualize a layout. Then, ask Gemini/ChatGPT to generate clean HTML and CSS code you can plug into your CMS or test in a sandbox.
You’re not replacing your dev or creative team. You’re giving them a prototype they can refine instead of starting from scratch. For most brands, this can save weeks of time – allowing you to go from idea → prototype → live LP in hours-to-days instead of weeks-to-months.
5. Ad Copy That Doesn’t Suck (And Doesn’t Take All Day)
Ad creative is the tip of the spear – it’s (often) the first impression your brand makes on your prospective customers/clients. Yet, somehow, most brands are still phoning it in: the same generic headlines and uninspired descriptions copy-and-pasted every ad group, the same dull landing page (hopefully you fixed that using the method in #4 above), and the same boring “Learn More” CTA.
The most common reason I hear for this? Most brands/agencies simply don’t have time to dedicate to more/better creative – so they settle.
It doesn’t have to be that way.
Start with this: ask ChatGPT or Gemini to generate headline and description options based on your landing page, offer, and customer pain points. You’ll likely need to fiddle with it to get something that you are happy with (remember - friction is good). Then, prompt it to create variants based on different personas or funnel stages (cold vs retargeting). Ask for headline length, tone (emotive vs benefit-driven), and even include performance tags like urgency or exclusivity. For RSAs, request 15-30 headlines and 4-6 long descriptions in character-limited format.
Then, call your custom Gems/GPTs and ask each one to rate/review each of those headlines/descriptions. For the poor-scoring ones, follow-up with suggested revisions or deletions. Once you’re happy with the output, load them up into the ad account.
6. Content Repurposing at Scale
One of the things I heard over and over from brands at SMX was how much more content they were trying to produce. At least half a dozen senior-level in-house marketers told me they doubled (or more) the volume of content they’re creating year-over-year.
That’s insane.
The reality? Most brands don’t need more content; they need more distribution. The same is true for most agencies (honestly, including ours): we have a TON of content, 99% of which is never used to its full potential.
Fortunately, there’s an easy win here. Let’s say you just wrote a blog post or recorded a podcast. Ask ChatGPT to turn that longform content into:
- 5 LinkedIn posts
- 3 Twitter threads
- A series of Quora answers
- One outline for a YouTube script
- Several short-form video hooks and CTAs
Then layer in visuals: use Midjourney to generate thumbnail or infographic ideas. Ask for social headlines and captions by platform. The result? 10+ pieces of channel-specific creative, all generated from a single piece of source material and completed in a matter of minutes.
Wondering which existing pieces of content to prioritize? Easy: (1) ask each of your custom GPTs/Gems to review your blog and select the 5 articles that resonate most with them; (2) have ChatGPT/Gemini conduct a gap analysis between your blog/content/resources and your competitors – (a) what do you have that they don’t? (b) What do they have that you don’t? (3) cross reference the list in (1) from the one from (2-a) – that’s where you start.
Bonus: everything in (2-b) is an opportunity for future content development.
7. Automated Audience Intelligence via Email Feedback Loops
Most feedback loops are either non-existent or fundamentally broken. Virtually every brand I talk to runs some kind of customer satisfaction survey once(ish) a year, if that – and the rest of the time, they’re relying on reviews (which are skewed toward the kind of people who leave reviews, which may or may not be your actual target audience/customers) and gut feelings from sales/customer support/customer success.
That’s not ideal. And it doesn’t have to be the way things are done.
Set up a simple process: send 2-5 personalized, plain-text emails per day from a senior exec (CEO, founder, etc.), asking recent customers/clients/prospects a few (2-3, but no more than 5) specific questions. Keep it simple and personal - no survey links or forms or anything. Just ask that they reply directly to the email. Use Gmail filters to tag the responses, then send them into Google Sheets via Zapier. From there, run Gemini/ChatGPT to classify and summarize key insights weekly.
This surfaces product issues, churn risks, and feature requests in real time, all while giving you the language your audience actually uses to describe their problems.
It’s fast. It’s scalable. And it makes your marketing more human.
The Bottom Line: The Brands That Win Are The Ones Who Do The Little Things
None of these 7 things are moonshots. They’re boring, simple, relatively basic workflows that capitalize on the strengths of AI right now AND give you (or your team) time back to do more important and more strategic marketing work.
None of these are sexy, but they work. So, do those. While everyone else gets frustrated trying to make the reality of AI match the hype, you can get ahead by playing to AI’s strengths today.
Skip the hype. Start building. The teams that win this year aren’t the ones who know the most about AI - they’re the ones who know how to use it. And if you want help building these workflows into your stack, drop a note/reply and we can connect.
Until next week,
Cheers,
Sam
Loving The Digital Download?
Share this Newsletter with a friend by visiting my public feed.
Follow Me on my Socials
