How to tell if AI advice is worth your attention

Before adopting the latest AI tactic, look for evidence, experience, and results you can replicate — and make sure the solution solves a real problem.

By Ryan Phelan,
Executive Leader, Email & Lifecycle Marketing
Published on October 6, 2026 • Last updated on October 6, 2026 • 8 minutes read

Share this article

Table of Contents

I love working with AI, but I need a breather. Partly because I’m up to my virtual elbows building an army of AI-driven minions (more on that later), but also because I’m exhausted by the constant stream of people telling me how to do AI right and articles promoting the latest AI tip of the day.

The truth is that much of this advice is, at best, questionable. It’s time for an AI reality check — not on its value and usability, but on how to evaluate the advice flooding our digital world and claiming to know the best ways to use it. After all, not every information source vets its authors or reviews its content as carefully as MarTech does.

Email marketers have been here before

Today’s AI environment has much in common with the early years of email automation. In those days, marketers eagerly discovered tools that would take email to its next level of value.

These tools would transform email marketing from simple one-off campaigns to personalized messages that merge with natural customer lifecycles through automated campaign-building processes. These initial automations performed the day-to-day grunt work and freed us to focus on more valuable pursuits. Sound familiar?

In the early days of email, marketers had few rules, little data, and no prebuilt, automated sequences. Teams made things up as they went along, trying everything and taking note of what worked.

Then we started talking to each other — at conferences, in discussion groups, on LISTSERV mailing lists, and at the occasional in-person get-together. Our own influencers emerged, amplifying their voices through email newsletters, blogs, conferences, and webinars.

They caught on because they shared what did and didn’t work for them, what they learned from testing, and what would scale across different verticals. A set of so-called best practices gradually emerged from these conversations.

As marketers gained experience with email, it became easier to discern whether someone was truly an expert or simply claiming the title and repeating what others had said.

That’s where we are now with AI and thought leadership. As a practical resource, AI has really only been available since 2021, dating back to the moment ChatGPT launched to the public. It wasn’t long before the flood of “do this, not that” advice on how to use it began.

But how many of these people were truly experts? How much of their advice was actually worth listening to?

3 questions for evaluating AI advice

Should you follow someone’s advice on AI or take it with the proverbial grain of salt? As you read advice from sources, look to see whether they (or their AI assistants) provide evidence or background, or just regurgitate what others have already said. The advice might not be wrong, but does it add enough unique value to merit your attention?

Here are three checks you can use:

1. Does the advice on AI wildly diverge from the consensus view?

In providing their advice, does the writer make claims that don’t align with the mainstream understanding of AI? Does the advice make you think, “Is this insane?” Or do you think, “This could work for me.” Does the writer veer dramatically from the view of other experts without explaining why?

Disagreeing with established AI best practices isn’t wrong by itself. It doesn’t mean that one person can’t come up with a singular great idea. But if it seems a little outlandish or if the author doesn’t show how someone else can replicate the same results, you should be wary.

This type of advice can also revive the “silver bullet” technology mirage that marketers have chased for years, spending time and money on things that don’t work in the long run.

When you read an article on AI and its uses, look for a groundswell of discussion or consensus on a particular tactic or way of doing things. Agreement isn’t groupthink — it’s the confidence of knowing that others can attest to the information’s value.

2. Does the expert show their work and how they got their results?

Marc is my really smart friend. He has built some amazing agentic AI models and written extensively about them. Through working with him, he’s become not just a friend but a mentor. While I’ve learned much from him, there’s one thing I haven’t yet learned: How he builds his models.

To understand and replicate his success, I need to know his process: where he started, what his query stream was, his conversations along the way, and his setups and failures. I see the results, and they are earth-shattering. But I don’t know how he got there, so I haven’t learned what I need to do on my own to achieve similar results.

That’s the part where an AI expert needs to show their work and their journey. It might not always be pretty, but it’s important information. It’s why I’m still trying to create my army of minions — my AI agents.

One day, I saw an Instagram ad for an AI dashboard that lets 20 agents make calls on someone’s behalf. You know, minions.

And I thought, “I want minions!” Setting up my own 15-agent system looked pretty easy. Three weeks later, I was still working on it. So I asked Marc for help, and he gave me a strategy and tactics to try.

I fed the information into my AI assistant, and it replied, “This won’t work because he uses a different model agent setup.” I’ll just have to save that “how I did it” post for another day.

When that article is written, I’ll tell you how cool the setup is, everything the system does, and what the output is. But I will also tell you how I struggled to build it, my experiences with GitHub, Trigger.dev, Supabase, Vercel, and other systems, and everything I learned from the experience.

The next time you read about someone’s amazing project, take note of whether they also show you the work behind it. The difficulty and effort behind what they achieved is key to determining if a similar approach makes sense for you.

AI runs on a token economy. When you spin up your preferred AI model, you’re spending both money and capacity. You must be sure that you’re spending your money, tokens, and time wisely and appropriately.

3. Understand the problem before implementing the solution

For any marketer considering a new technology or workflow, it’s essential to think about strategy before tactics. In other words, know the “why” of strategy before you invest in the “how” of tactics.

Don’t start with “Let’s build it.” Begin with “What problem do we need to solve?”

Ask your AI assistant questions like the following:

  • Does this make sense for my business?
  • Is this realistic?
  • Will it work?
  • Will this accomplish a goal or solve a problem?
  • How can I make money with this?
  • Is the outcome worth the work?

Anybody can claim to be an expert and describe some AI-powered system. You must read skeptically and look for information that answers these questions.

The evolution of the abandoned-cart workflow is the gold-standard example of this and was once a hot topic among email marketers. Teams had the technology to send automated email reminders to customers who left items in their carts, but many had to wait 24 hours or longer to get the necessary data to trigger the email. That delay often killed the chance to bring the customer back.

Then, in 2010, Charles Nicholls of SeeWhy shared data and advice that helped marketers rebuild their cart programs and capture more sales. For example, the data showed that the best time to send an email reminder was an hour after a sale. That gave marketers the business case to demand data sooner.

This doesn’t mean that an hour later is the best time to send follow-up emails to everybody. But it provides a valid starting point for testing, rather than simply guessing.

The same is true of the growing list of AI best practices. Someone can claim that a certain prompt or query stream is the best way, but that doesn’t mean it will work for your brand, customers, operations, or market.

Who is a thought leader in AI?

There’s a lot of noise around AI today, and I may have added to it here. But I also hope I’ve helped you read advice more skeptically. As with everything in thought leadership, people can call themselves experts, but that doesn’t mean they are.

What makes a thought leader deserve the title and your trust? In addition to my three test questions, I would consider their background, work experience, other publications, and brands.

Also consider how much they give back to their industry. Do they share their advice for free on podcasts and webinars, as keynote speakers, as authors, and as mentors?

True thought leaders aren’t just smart people. They’re people willing to share what they know to help lift someone else’s boat.

My greatest concern about this new generation of AI thought leaders is that we need to be cautious about whom we listen to. Not to stymie innovation, but to be realistic about its prospects.