AI Literacy Is Not Instant Literacy. Ann Handley says Reading and Judging is Judging
Ann Handley posted something on LinkedIn last week that stopped me in mid-scrolling. He is a Wall Street Journal best-selling author and one of the most respected voices in marketing, and wrote:
“AI literacy is not instant literacy.
His post went on to ask a question that no one in the AI training industry seems to have: “Why do we keep teaching people how to use AI – without teaching them when not to use it?”
I sent him a message. I had to know where someone would go to learn that.
His honest answer: “I don’t know of a course that only teaches this. At MarketingProfs, our sessions on AI usually include a few slides on when not to use AI, or how to protect against hallucinations, but I don’t know of an entire session or series.”
He added, “I think that’s really the story, and why I wrote what I wrote. We have an entire industry built around AI skills training — rapid developer bootcamps, certification programs, tool tutorials, a million LinkedIn posts about the perfect information you need to do this or that or even if you’re behind. What we don’t have is anything that asks: What do you have to give up when you use a tool?
That gap is real, and more important than the AI training industry is currently admitting.
Prompt Literacy Takes Afternoon. Learning and Judging Takes Years
The differences that Ann draws are not subtle once you see them. Reading and writing quickly can be taught in the afternoon. You learn syntax, structure, iterative refinement loop. You learn to specify, to add constraints, to tell the model what not to do and what to do. This is really useful and a quick read.
Judging literacy is something else entirely. Knowing when the speed of AI output is actually eroding is something you needed to build slowly. Seeing where the struggle itself is the point, where the conflict of not knowing the answer is what produces the expertise that will be important later. It’s understandable, as Ann put it, “when AI helps and when it cuts through the very struggle that teaches us something.”
One commenter on his post put it exactly:
“Reading and writing quickly can be taught in an afternoon, while learning and judging takes years, because judging most of the time is knowing the importance of the struggle you will be going through.”
I have been teaching an online course on AI content to a truly trusted audience for several years. And I’ve spent the last few months analyzing what the field of AI training really offers doctors. The pattern is consistent. The courses available (and now there are many of them) teach you what tools can be made. The best ones teach you how to use them wisely. Almost none of them teach you when to put yourself down.
This is not a small gap in the learning process. It is an important question of the present time.
Why the Gap Exists
The AI training industry has an incentive structure problem. Courses that teach you how to use tools create the need for more tools, more courses, more certifications. There is no business model to prevent teaching. No one builds a fast engineering bootcamp whose main lesson is “sometimes they don’t.”
But the costs of skipping a judicial question are real and measurable. Anthropic’s own research found that junior engineers who relied heavily on AI coding agents showed poorer understanding of their work when tested later. While the tool produced a result, their struggle to create technology did not materialize. Output and expertise are not the same thing.
For SEO professionals and content marketers specifically, exposure is straightforward. MIT’s AI Labor Exposure Map, which I wrote about last week, found that nearly three-quarters of the time a marketing professional spends at work goes to tasks that AI can handle. The question is not whether AI should be used for those tasks. For many of them, it should. The question is which jobs in the 74% are actually where doing the learning, where the outsourcing also brings out the understanding you needed to build.
That question requires judgment. It cannot be answered by request.
Tradition, Not Academic Work
When I asked Ann where doctors should go to improve this judgment, her second message was to reframe the question entirely.
“Do we really need a course? What we need instead is approval and better modeling. Leaders who clearly choose the long road. Managers who say loudly if they won’t use AI for certain things, here’s why. People see the value. They say another way: culture not academic work.”
That reframe is worth living with. Judging about when not to use AI is not a skill that is imparted through a certification program. It’s a professional practice that is passed on by watching, watching someone you respect choose to do something slowly, pushing people into the dark, and then explaining why.
Ann has a book coming out in February 2027 from Penguin Random House called “ASAP (Aslow As Possible): When to Take the Long Road in the Short Country.” The title captures the tension well. In a professional culture that has made speed the ultimate virtue, choosing to slow down requires not just judgment but courage: being willing to stand out takes longer when everyone around you is speeding up.
What Employees Can Actually Try Right Now
Ann’s point about cultures instead of subjects is correct in the long run. But while that culture is being built, doctors need something tangible. Here’s a repeatable workflow, taken from a test I did with the editorial team at The Acton Exchange, a nonprofit community newspaper in Acton, Massachusetts, in November 2025.
The team faced a deadline problem. The steering committee had just held a three-hour work session on a key school district restructuring question, reviewing 156 pages of literature. The meeting was not recorded, meaning no transcript is available. But 101 pages of additional information and 55 pages of public comments that the committee received early are accessible.
So, the group tried something new. We have created detailed information that specifies what the article needs to achieve: accurate and reliable information, an engaging story, relevant to residents. We loaded all 156 pages into four AI engines simultaneously: ChatGPT, Gemini, Perplexity, and NotebookLM. Each engine took a different route to the same information and the same source material. ChatGPT generated 748 words focused on data and process. Gemini produced 712 words focusing on why the status quo is no longer valid. The confusion generated 1,232 words focusing on what the options mean for residents. NotebookLM generated 1,506 words organized around five amazing facts.
We reviewed all four drafts together in an all-hands planning meeting. The confusion draft was very accurate and very useful as a basis. We chose it as our starting point. Then we did what no AI engine could do: We added direct quotes from the people in the room, reflecting the voices of the community Acton Exchange exists to represent.
The important lesson from this test is not which engine performed best. That is the revealed process of judgment. City Manager John Mangiaratti noted a few weeks ago that the tools were useful for the first 75% of content, but that “the remaining 25% of detail, variation, and context are irrelevant or wrong.” Superintendent Peter Light agreed, adding that quality is improving with better installation instructions.
That 75/25 split is a viable framework for any content workflow. Use AI to get 75% of the way there faster. Then use human expertise, primary source verification, and direct observation to close the gap. 25% need someone is not a disruption in the workflow. This is where judgment resides.
Before adopting any AI tool into your content plan, have a clear discussion with your editor or team about which tasks the AI will handle and which require human supervision. Write your information. Use the same information with more than one engine when the stakes are high. Validate your output against primary sources before publishing. And disclose your process to your audience, as Acton Exchange did under this published article.
Ann Handley is right that the real skill is judgment: knowing when speed is useful and when it actually destroys something you needed to build. The Acton Exchange test did not resolve that question. It made the question tangible in a way that a quick engineering course never could.
Being able to read and write quickly gets you to 75%. Learning and judgment is what closes the rest.
Additional resources
Featured image: Yuriy2012/Shutterstock
