Code as Content: Why I Ship Tools, Not Articles

There’s an idea that’s been rattling around my head for more than a decade. I’m pretty sure I first heard Rand Fishkin call it “code as content.” Though I might be misremembering, so don’t quote me on the attribution.

The idea is simple: the most durable content you can publish isn’t an article. It’s a tool.

For years that was just a nice edge. Now it’s becoming survival, because the ground under content marketing just cracked open and most people are still writing blog posts on top of it.

The uncomfortable truth: text is now a commodity

Nobody wants to say this out loud. That “educational content” strategy we all worshipped for a decade, the one I used to grow the AdEspresso blog to millions of visits, is now mostly free labor for Google and the LLMs.

You research it, you write it, you publish it… and an AI summary swallows it whole and hands the answer straight to the user. No click. No visit. Just you, feeding the machine that replaced you.

Educational content is the casualty

This isn’t a vibe. It’s measurable, and the numbers are ugly.

Pew Research ran an independent study, no vendor with an axe to grind. When Google shows an AI summary, people click a traditional result 8% of the time, versus 15% without one. Roughly half the clicks, gone. Clicks on the links inside the summary? 1%.

And guess which queries trigger those AI Overviews? Ahrefs found 99.2% of them are informational (source). Informational. As in the exact “how to,” “what is,” “best way to” content the whole inbound playbook is built on.

The how-to article isn’t a growth engine anymore. It’s a donation to Google.

“Education isn’t enough” — and it never fully was

Ryan Law at Ahrefs put it best: the marginal cost of content has collapsed to roughly zero. When anyone can spin up a competent 2,000-word guide with two prompts, explaining something is no longer a moat. Information arbitrage is over.

I lived the good years of that arbitrage. I’m telling you they’re not coming back.

Two ways out of the same trap

Now, I’m not saying content is dead. That’s the lazy headline. What died is undifferentiated text. There are two real ways out of the commodity trap:

1. Radically human content. Opinion, story, proprietary data, a take an AI can’t fake. (Ryan Law’s bet.)
2. Code as content. Ship a tool instead of an article. Make the page do something.

And here’s a fun U-turn on myself. Years ago I wrote that Google killed short-form content, and that it was bad for everyone. I still think it was… but in a zero-click world, where SEO traffic to articles is drying up anyway, short and sharp and contrarian POV pieces make sense again. Just don’t write them for Google. Put them where the audience actually is: LinkedIn, X, a newsletter. That’s content as a point of view, not content as an SEO play.

This post, though, is about the second exit. It’s the one I keep coming back to, and the one I’ve watched work with my own eyes.

Why the tool wins in the AI era

The tool is the content

Eli Schwartz, who literally wrote Product-Led SEO, nails it: an AI can generate infinite text, but it can’t replicate your utility or your proprietary dataset. The commodity is the text. The scarcity is the function.

So stop asking “what article should I write to rank?” and start asking “what job can I do on the page?”

It’s zero-click content, taken to its logical extreme

Everyone’s panicking about zero-click search. A good tool flips it into an advantage: the visit delivers value without the user leaving, and the action converts right there.

Look at HubSpot’s Website Grader. One free tool, launched back in 2007, that grades your site in five seconds. By 2010 it had graded over 2 million sites. (You’ll see people claim it drove “10 million leads.” I couldn’t find a real source for that, so I won’t repeat it. The 2 million is documented, and it’s plenty.) The click you’d have lost on an article becomes an action you captured on a tool.

And this isn’t some relic from 2007. A few weeks ago Dharmesh Shah, the HubSpot cofounder who built the original Website Grader, shipped a new free tool called HubGrader. It hit 10,000 users in two weeks. Nineteen years apart, same exact move. The free-tool playbook doesn’t age.

The proof: free tools that are quietly winning

None of this is theory. Here’s a spread of free tools, all free to use and monetized elsewhere, putting up numbers no blog post in their niche could touch.

Finance calculators are the clearest case, because the tool sits right on top of real money intent:

• Zillow’s Zestimate. A free instant home-value estimate on ~104 million houses. Zillow pulled 221 million monthly users and $2.6B in revenue in FY2025 (source). The estimate is free; the mortgages and agent referrals are not. The tool is the funnel, at national scale.
• NerdWallet. Free calculators and comparison tools. $836.6M in revenue in FY2025, up 22%, and that’s from an audited SEC filing, not a marketing deck. Their own 10-K names organic search as the growth engine.
• SmartAsset. Free tax and retirement calculators feeding an advisor-matching marketplace. A $1B valuation, built on “let me just calculate my paycheck real quick.”

Utilities win because the tool is the action, and because they’re wildly linkable:

• TinyPNG. Compress an image, done. That one utility has earned 42,700 referring domains and ~5M visits a month (Semrush). Forty-two thousand domains linking to an image compressor. That’s the whole thesis in one stat.
• Smallpdf. Free PDF tools in the browser. 40M+ monthly users, and a $30M acquisition to show for it.
• Omni Calculator. 3,700+ free calculators, bootstrapped, zero VC, monetized purely with ads, grown from 20,000 to 18M+ visits a month. The purest “the tool is the business” play I know.

And the developer world runs on this too:

• JSON Crack. Paste JSON, see it as an interactive graph. 44,000+ GitHub stars, embeddable via npm, open source. No article will ever substitute for “paste it and watch it render.”

One more, in a different vertical. Tax Democracy is an Italian tax simulator that sums every income source into one progressive calculation and shows you the step-by-step math. Its edge is exactly the Schwartz point: the logic is proprietary and the user takes a real action. And it’s built as a constellation of micro-tools, net salary, child benefit, and so on, each one its own landing page. Those micro-tools rank really well and keep growing, and they’re the main lead driver into the full simulator.

Here’s what two of them look like. Each is a standalone page, ranking on its own query, feeding the same funnel:

But let’s be honest: tools aren’t magic

This is where I lose the hype crowd. A tool does not rank itself. Anyone who tells you “just build a calculator and the traffic comes” is selling you a course.

A tool is a multiplier, not a shortcut. What makes it a great one is that, unlike a blog post, you can promote it like a product. Let me show you the mechanics with tools I actually shipped. (Some don’t even exist anymore. Doesn’t matter, the how is the lesson.)

You can run a product launch

You can’t launch a blog post on Product Hunt. You can launch a tool. Email lists, Product Hunt, communities, press… a free tool earns an event’s worth of attention that an article never will.

Every tool we put out at DivByZero went up on Product Hunt, and most landed in the top 3. SaaS Examples hit #2 of the day and #2 of the week with 485 upvotes back in 2022. Try getting that kind of launch moment for your latest listicle. And it compounds: free tools ended up driving a big slice of this blog’s traffic too. I broke down the numbers in Road to a 1 Million-visit blog.

A good tool is a backlink magnet

This is the one that changed the game for me. At AdEspresso we built an ads gallery, a searchable library of Facebook ads by theme, keyword, advertiser. This was before the ad transparency era, so nothing like it existed.

It was the best marketing investment we ever made. It generated hundreds, probably thousands, of links. It lifted the backlink profile of the entire domain and became a flywheel that made our already-strong blog rank even better.

A well-made tool, one that isn’t trivial to clone and leans on proprietary data, is one of the most linkable things you can put on the internet.

Tools sneak past influencer neutrality

Here’s a subtle one most people miss. Influencers like to stay neutral. They sell courses, they chase the big audience, and they don’t want to tie themselves to one vendor. So they won’t push your product, no matter how good your relationship is.

At AdEspresso, the big names in Facebook advertising happily talked about the Facebook Ads Manager all day… but they wouldn’t push AdEspresso. Fair enough.

Free tools broke that wall. We released Pixel Caffeine, a free WordPress plugin to install the Facebook Pixel and build custom audiences. It didn’t compete with anything they were promoting. It was just useful. So every major influencer in the space linked it and mentioned it. For free.

…and yes, they might rank on their own too

One honest caveat in the other direction. Promotion is necessary, but it’s not impossible that a tool earns organic ranking on its own merits.

Koray Tuğberk GÜBÜR has an interesting hypothesis: Google may be learning to recognize what is and isn’t a tool, and what it can and can’t embed inside an AI summary. A tool the AI can’t just execute inline forces the click to your page.

Is that a law? No. It’s a hypothesis, and ranking still depends on a hundred other things: your backlink profile, your domain authority, all of it. Being a tool, by itself, isn’t enough to rank. Same as it isn’t enough to get promoted. But the direction is worth watching.

The catch: build a moat, or die like the others

Here’s the flip side. A tool with no moat gets commoditized just as fast as an article. And lately, the AI eats it alive.

Ask Chegg. Homework answers behind a paywall, fed by Google search. Market cap down ~99% from its peak, gutted by ChatGPT and AI Overviews. Or Stack Overflow, where question volume has dropped 76% since ChatGPT launched. When your “tool” just returns an answer the LLM can now give inline, you’re toast.

So the moat is everything. In the AI era, with Claude Code and Cursor, a tool is easy to build, which means if all you have is a good idea, someone clones it by Friday. What actually defends you:

• Proprietary data. The one thing an LLM can’t regenerate and a competitor can’t copy. (Zillow’s Zestimate. Ahrefs’ index.)
• A real action, not a fact lookup. If the AI can compute your answer in the summary, it will. Make the user do something (upload, simulate, generate) and hand back a personalized or downloadable output.
• Linkability. Embeddable, genuinely useful, worth citing. That’s how you earn the distribution you can’t buy.

A “percentage calculator” is a fact the AI will just do for you. A multi-variable simulator running on data you own is not. Know the difference.

Tools are the new lead gen

One more reason I keep betting on this. Lead generation through content is dying. Webinars, ebooks, newsletters… the conversion just isn’t there anymore. (I’ve generated over a million leads across three SaaS companies, so I’m not saying this from the cheap seats.)

A well-made tool where part of the value unlocks with a signup still works beautifully. The user already did the work and already got value, and handing over an email to unlock the rest feels like a fair trade, not a toll. Gate the export, gate the saved results, gate the advanced scenario. That’s your funnel: the same product-led logic that ate SaaS, pointed at your top of funnel.

The part nobody says out loud: brand vs. creator

Let me end on a take I haven’t seen many people make.

Feeding the LLMs is not equally bad for everyone. It splits sharply in two:

• If you’re a brand or a product, it can actually be good. You stay top of mind. If ChatGPT names you as the best tool for the job, fewer people show up, but the ones who do convert at an insane rate. You traded volume for intent, and intent pays.
• If you’re a pure content creator, it’s brutal. No click means no visit, no ad impression, no way to eat. You did the work; the AI monetized the attention.

That’s the whole argument, really. In a world where the answer lives in the AI summary, you want to be the thing the answer points to (a tool, a product), not the text the answer is made of.

Be code, not content.

That’s it. That’s the bet. Stop writing the 2,000th article on a topic and go ship the one tool that solves it. Just remember to promote the hell out of it… and to build something they can’t clone by Friday.