How I Use AI at Ask Leo!

It’s my fast (but unreliable) intern.

You may have noticed a small update at the top of every page at Ask Leo!. Right above the search box on the right-hand side, it now says “Written by a real human.” That real human, of course, would be me. I suspect this will set me apart from future waves of AI-generated content.

But that doesn’t mean I avoid AI completely. I want to share how AI helps me and how it sometimes misses the mark.

And since AI is changing so quickly, I’m updating this article a little sooner than most. Much has changed in two years.

AI at Ask Leo!

I write every Ask Leo! article myself, but I do use AI as a tool along the way. It helps with ideas, summaries, and images, but rarely gets things exactly right. I think of it as a fast but unreliable intern: it’s useful, but its work still needs careful review.

Using AI at Ask Leo!

I write every article on Ask Leo! I don’t accept guest posts, sponsored posts, or anything similar. In some ways, it limits my reach, but it’s one of the ways I’ve chosen to set myself apart from other sites.

Of course, I use a variety of tools to help me. AI is just one of those tools. (I also have a fantastic human editor who continues to make me look like a better writer than I am.)

AI is incredibly powerful, sometimes unreliable, and in constant flux.

And it’s something I’m having a lot of fun with.

In this article, I’ll describe the AI processes I used for the immediately preceding article, Is It Safe to Let Your Browser Remember Passwords? as an example. Note that since that article is finished, some of the AI responses here might differ from the published article. Such is the nature of AI: it includes a certain amount of randomness in its responses.

These days, I use mostly Claude, with Google’s Gemini Nano Banana Pro for images. Naturally, that could change over time1.

I use these tools for distinct tasks:

• Ideation
• Summarization
• Image generation

Ideation

Even though I already knew my answer to the question, I wanted to see if AI would derive the same solution.

Prompt

Is It Safe to Let Your Browser Remember Passwords?

AI response

Great topic for your Ask Leo! audience2. Here’s a straightforward breakdown:

Is It Safe to Let Your Browser Remember Passwords?

The short answer: It’s better than nothing, but it’s not the best option.

What Browser Password Managers Do Well

Every major browser — Chrome, Edge, Firefox, Safari — has a built-in password manager that offers to save your login info and fill it in automatically. For people who would otherwise reuse the same password everywhere or write passwords on sticky notes, this is a real improvement. Research analyzing nearly 19 billion leaked passwords found that 94% were reused or duplicated — so anything that encourages unique passwords per site is a step in the right direction.

The Real Risks

Browser password managers lean toward convenience over security, and that creates some specific problems:

1. Weak protection on your computer While browser password managers store passwords in encrypted databases, they store the associated encryption keys in predictable locations — it’s comparable to locking your house but leaving the key under the doormat. If malware gets on your computer, your passwords can often be extracted quickly.

2. No master password by default Chrome and most browser password managers use your OS login as the decryption key. If you’re logged into Windows or macOS, anyone — or any malware — with access can decrypt your password database. There’s no additional master password protecting it.

3. Physical access is a problem Browser password managers won’t generally ask you to regularly re-authenticate yourself, meaning that if someone steals your laptop they will have easy access to all your passwords.

4. Everything’s connected to one account If you’re using Chrome, everything is tied to your Google account — your history, passwords, cookies, and account settings. If someone can access your browser or the account you use in your browser, they can open up everything.

5. Vulnerable to malware and extensions Cybercriminals use malware to extract saved passwords from browsers. Once installed, it can scan the browser’s storage for saved passwords and send them to the attacker. Earlier in 2025, Google Chrome leaked passwords through an extension-based attack.

The Better Option: A Dedicated Password Manager

Security experts consistently recommend standalone password managers — like Bitwarden (free), 1Password, or NordPass — over browser-based ones. Here’s why they’re stronger:

• • They use a master password that only you know, adding a real lock on top of your vault.
  • They use strong encryption (AES-256) and many use a zero-knowledge design, meaning even the company can’t see your passwords.
  • They work across all browsers and devices, not just one browser.
  • Many offer breach monitoring — they’ll warn you if a site you use has been hacked.
  • They offer stronger security features; the main risks come from phishing, weak master passwords, or faulty updates — not from the encryption being cracked.

Bottom Line for Everyday Users

Using your browser to remember passwords is better than nothing, but a dedicated password manager gives you meaningfully stronger protection — especially for your most important accounts.

Sources: Keeper Security, Bitwarden, The Cloud Standard, Chapman University CISO Blog, IT Foundations, Spin.AI, Sticky Password

Reaction

That’s a pretty decent answer. There are minor points I might quibble with; for example, breach monitoring is nice, but not directly related to the question and not something I consider when using a password manager, so not something I’d complicate an answer with. And I would write (and did write) the answer in a slightly different way. But that was a solid answer; I wouldn’t feel bad if someone used Claude to ask this question and walked away with this as their answer.

This would have helped me identify and cover most of the important bases, if I didn’t already have them in mind. (Which I did.)

Writing

So, I wrote the article — my article, from scratch, without any help other than spellcheck. (I did not have Claude’s answer above in front of me; I generated that for fun for this article.)

There are two parts of every article I often use AI to draft:

• The summary
• The excerpt

And of course, my editor works her magic at some point as well.

The summary

AI is good at summarization. It’s not uncommon for it to pick up on things that my manual summarization might not capture or emphasize appropriately.

My prompt

Please summarize in 50 words or fewer, in a style similar to the article itself, using a 6th grade vocabulary and the second person:

This is followed by the full text of the article as written so far.

AI response

Your browser can remember passwords, but it’s risky. Anyone who sits down at your computer can see them in just a few clicks. A dedicated password manager is safer, more secure, and works across all your browsers. It’s the smarter way to protect your accounts.

My reaction

It’s not bad, but it’s not great. And while it’s close, it’s not in my writing style. The approach I use, if this initial response feels “close”, as this one does, is to copy/paste it into my article’s summary and then edit the heck out of it. I’d probably use something like:

Your browser can remember passwords, but it can be risky. For example, the default configuration often allows anyone with access to your computer to see them in just a few clicks. A dedicated password manager is safer, more secure, and works across all your browsers.

There are some pedantic changes:

• “It’s risky” became “it can be risky”. The former is not an absolute.
• “Anyone who sits down” became “anyone with access”, with the additional qualifier that it’s the default configuration at play.
• I removed “It’s the smarter way to protect your accounts” because it’s a value judgment (I try to make the summary close to the answer and without additional judgment), and doesn’t add value to the summary.

These kinds of transformations are common when I turn an AI-generated summary into something I’m more comfortable with.

The excerpt

The excerpt — that box of text above the image — is used in a couple of places, depending on how the article is accessed.

My prompt

Please write a paragraph ‘tease’ of no more than 50 words for this article that mentions what it’s about, and the information it’s about to deliver, in a way that will encourage individuals to read it.

AI response

Your browser offers to save your passwords — but should you let it? Before you click “Save,” you need to know how easy it is for anyone to see those passwords, and why a dedicated password manager is the safer, smarter choice for protecting your accounts.

My reaction

Again, a fine result, but not quite me. Here’s what I’d turn that into.

Your browser offers to save your passwords, but should you let it? Before you do, let me show you how easy it is for anyone to see those passwords and why a dedicated password manager is a better choice for protecting all your accounts.

The changes include:

• Replacing the em dash (–) with a comma. There are conflicting reports as to whether em dash is overused by AI, and perhaps even a signal that AI has been used, but in most cases, it’s just not needed.
• “Save” replaced by “do”. Not all browsers use the literal “Save”.
• “You need to know” replaced by “let me show you”. “You need to know” seems too formulaic and almost demanding. “Let me show you” reinforces the idea that there’s a real person behind this article.
• “Safer, smarter” choice replaced with “better”. I dislike “smarter” in situations like this, since it implies not making this choice means the reader is less smart, which isn’t the case and something I’d never want to imply.

Once again, these kinds of transformations are common for me.

You might ask, “Why use AI at all if I’m making all these changes?” Fair point. To that point, the published example article didn’t use this technique: the summary and excerpt were 100% Leo-created. They’re OK (which is why I left them when I updated that article). I find that the AI approach gets me somewhat better results. If nothing else, it gives me a starting point from which to craft my own version and may save me some time.

Image ideas

I’m not a very creative person, at least not when it comes to images and illustrations. I’ve been using AI to help me out there for quite a while now. It’s way more “creative” than I am.3

There are two ways I generally have AI help me: I ask it to create an image based on the article text, or ask it for a list of ideas. Let’s start with the latter.

My prompt

Please describe 5 bright and engaging hero images for the article.

AI response

Here are five hero image concepts for the article:

1. 1. The Open Vault vs. The Locked Vault — Side-by-side comparison: a wide-open browser window with passwords floating out visibly on one side, and a locked, glowing vault on the other. Bright, clean colors with a clear visual contrast between “unsafe” and “safe.”
   2. The Sticky Note Nightmare — A desktop computer completely covered in colorful sticky notes, each with a fake password written on it. A web browser on screen offers to “Save Password.” Chaotic but humorous, with bright colors making it eye-catching.
   3. The Unlocked Door — A giant browser icon serves as a door standing wide open, with password symbols and login credentials spilling out onto the floor. A determined-looking cartoon hacker peeks around the corner. Bright, slightly playful style.
   4. The Password Parade — A cheerful but alarming image of passwords marching in a line out of a browser window and right into the hands of a shadowy figure. Bold colors and a slightly comic style keep it engaging rather than scary.
   5. The Two Roads — A split image showing two paths: one leading to a browser with an unlocked padlock and visible passwords, the other leading to a glowing, secure password vault with a strong padlock. A person stands at the fork deciding. Clean, bright, optimistic colors.

My reaction

Those are better ideas than I would have come up with on my own, for sure. When I’ve done this for other articles, I’m often drawn to one of the prompts, but then tweak it a little before asking AI to create it.

In this case, however, #2 turns out to be pretty darn close to what I had already created.

Image generation

These days, I use Google’s Gemini and its oddly named “Nano Banana” to generate images.

Prompt

Please create a 16:9 image in pixar style of a desktop computer covered in sticky notes each with a password on it. On the screen is a web browser offering to save a password. On the floor is a Corgi with a sticky note on its nose which also has a password on it. The Corgi is looking cross-eyed at the note.

AI response

Reaction

Not bad at all. My only quibble? I address mostly Windows and PC questions, and AI image generation seems drawn to Apple products. (That’s the Mac OS interface with the Safari browser on screen).

There are two technical things wrong with the image, though: while I asked for 16:9, I actually want 2400 x 1260, the image dimensions that display the best without cutting things off on Facebook when the article is shared. The other is the Gemini branding star in the lower right.

I use AI in either Canva or Adobe Photoshop to erase the branding and perform “generative expand” on the image to make it a little wider to the dimensions I want. At the same time, I put a black border on it. Here’s the result.

I generally put the original prompt as the title attribute on the image for screen readers and the like.

Postscript

I used portions of my process above on the article you’re reading now.

Here’s AI’s un-edited summary:

Leo writes every Ask Leo! article himself, but uses AI as a tool along the way. It helps with ideas, summaries, and images — but rarely gets things exactly right. Think of it as a fast but unreliable intern: useful, but everything still needs your careful review.

Its concept for an excerpt:

You’ve probably wondered whether AI writes the articles you read online — including this one. The answer at Ask Leo! is no, but AI does play a role. Here’s an honest, behind-the-scenes look at exactly how AI helps, where it falls short, and what a real human still has to do.

I asked Claude for ideas, and it came up with the following. I lightly edited it and passed it on to Gemini.

Please create a 16:9 image in pixar style of The Helpful Intern — A bright, cheerful illustration of a friendly robot sitting at a small desk next to a human writer, handing over a sticky note with ideas on it, while the older male human works at a larger desk. The human’s computer screen displays “Ask Leo!”. There is a Corgi curled up sleeping in a dog bed against the back wall.