AI Didn't Kill Writing: How a $7K/Month Blog Writing Hustle Still Grows

The episode opens on a single chart. Chris pulls up Google Trends for the phrase 'hire a writer' over five years: a visible dip around the arrival of ChatGPT in 2023, then a steady, sharp climb over the next two years. The takeaway is not that writing vanished, but that good writing became more selective and valuable. Companies stopped wanting hundreds of interchangeable posts and started paying $900 to $1,200 for pieces where the author has actually clicked through the product, felt the confusing step, and can translate that hesitation into clear prose.

Why Demand Ticked Back Up

Guest Steph Spector gives that chart a human face. Based in Austin and writing since 2016, she edited sales-enablement materials for Apple around the time ChatGPT appeared, then wrote full-time at Zapier before going freelance three weeks ago. Today she ships 5 to 10 blog posts a month, charging $800 to $1,200 per piece, each taking five to six hours, with additional work coaching executives and helping high schoolers with college essays. The blended total lands near $6,000 to $7,000 a month, built on writing, editing and coaching in one family of skills.

Her core line is that writing is thinking. For a tutorial on a tricky automation, she hits record on a Loom-style capture, narrates as she builds, and makes the uncertainty explicit: this feels fiddly, what if I click here, why is this label unclear. Capturing that raw reaction is the gold she feeds the machine. It preserves the difficult bits that would otherwise take hours to reconstruct from memory and gives the later draft a spine of real experience.

Writing Is Thinking: Record, Then Route

She does not bet on a single model. Gemini is strong at turning video and audio into structured notes and noticing hover-pauses that signal confusion; Claude is better at shaping marketing copy in her voice; ChatGPT helps with research and drafting. She routes them through Cursor, which can send different steps to different models in one flow. In practice, Gemini watches the screen recording and extracts sticky points, then Claude takes those notes plus a template she designed for that tutorial and produces a solid first draft that she edits for rhythm and personality.

Underneath sits a style guide and template discipline. At Zapier she learned to feed the model the house rules — capitalization, bold and italic conventions, banned words, formatting — and to give it structural templates for each content type so the model fills carefully bounded blanks. At Apple, AI use was not permitted during her contract period, so she matured this practice at Zapier, where teaching the model the expected shape of the output made drafts land closer to publishable.

The client pipeline leans on Upwork, where she started in 2016 and still works. When she looks for work, she applies to about ten hand-picked jobs a day, favors postings likely to repeat, and expands the contract by offering to upload the post herself. A good working relationship turns a one-off website rewrite into years of retained work, which is why she sees the marketplace as lead generation rather than a race to the bottom. At Zapier she saw the same dynamic from the other side: teams hold on tightly to reliable freelancers.

From Upwork to B2B SaaS: How the Business Works

Her current briefs are mostly B2B SaaS blogs of 2,500 to 3,000 words for firms like Zapier, JotForm and Softr: versus posts such as Claude Code against Cursor, eight alternatives to QuickBase, or step-by-step build guides for a specific workflow. Clients usually supply target keywords, so she does not need to be an SEO technician, but she must know heading hierarchy, appropriate length, clean titles and how to avoid stuffing. A newer layer is GEO — generative engine optimization — making content citable inside AI answer engines.

Why five to six hours if AI helps? Most of that time is hands-on product time. She signs up, clicks around, takes usable screenshots, then reads Reddit threads and community chatter to see where people actually struggle — a confusing pricing tier, a brittle integration — and verifies it herself. Trying to automate the whole pipeline top to bottom is the classic trap she sees in other writers: endless configuration, upkeep, middling output, then burnout she calls AI brain fry. The better rule is to use AI where it hurts most and keep the rest deliberately light.

Enterprise hiring signals reinforce the point. Listings from Microsoft, Apple, Lovable and Anthropic for lead writers now ask for grammar and storytelling plus fluency in meme culture and a personal creative pursuit on the side, like playing an instrument or running a newsletter. Hiring managers want a well-rounded person who can import lived texture into a draft, because at a moment when any YouTube title like 'seven best side hustles for 2026' can be read off a synthesized script, the human story — in Chris's case, a grass-fed beef venture where a cow's leg punched through a trailer floor — is what registers.

Why Human Stories Outlast Watermarks

The strongest evidence is a natural experiment that played out over the last year. Many companies tried to go fully AI for content, freelancers heard 'we are moving on,' and about a year later the same companies circled back with 'are you free to help with this?' Platform moves sharpen that shift. Snapchat signaling no more purely AI videos, YouTube pushing faceless AI channels down the recommendation stack, and the story of Sora — an AI-only feed that surged to about 100 million users then shut down months later — all point to audiences craving real material. Anthropic's invisible marking of Claude outputs, with its undisclosed algorithm and nascent removal tools, is the infrastructure version of that same negotiation.

Steph puts numbers on her own mix: 70 to 80 percent of the words are hers, structure and thesis close to 100 percent hers. Zapier's AI Workflow Index frames the broader pattern: even among the most AI-adopting mid-market and enterprise teams, only about 18 percent of steps in so-called AI workflows are actually AI; the rest is rules, logic and app connections, with a human approval step before anything ships. That is why feeding 100 articles into a Claude Project still does not replicate voice. The model can mimic cadence but does not carry childhood memories or the way a writer's thinking evolves through doing the work; freezing the voice in a prompt keeps it static while a practicing writer keeps moving.

From a buyer's perspective the calculus is speed plus trust. At Zapier she saw freelancers win on turnaround when a timely topic lands — say a tool is briefly discontinued and search interest spikes the same day. Being the credible person who publishes that afternoon earns both rankings and thought leadership. A creative-writing background shows up not as ornament but as control over sentence rhythm and engagement: when a technical section threatens to sag, she has a thousand small moves to bring it back alive.

A 90-Day Playbook Without a Degree

For newcomers, Stephanie's late-night research is useful. To prep for the episode she hired four or five writers on Upwork for 20-minute chats, got 15 thoughtful applications in two days, and talked at length even though she paid for 20 minutes. None had a writing degree. Their consensus was a three-part 90-day arc. Month one: map the landscape — books and fiction on one side, corporate and marketing copy on the other — and choose a lane that tugs at genuine curiosity rather than a generic 'freelance writing' label.

Month two is niche focus. Going broad early slowed Stephanie down; specializing in one form built trust. Her barbecue example travels well: a note to a small beef-protein powder brand about a flavor she liked, followed by 'I see you have a blog — need a writer?' turned into a paid post without ever asking to sell. The same pattern showed up at her friend's e-commerce ad agency, where a well-branded Shopify store discovered on Instagram swipes led to a direct message and a client.

Month three is proof plus outreach craft. No credential is required, but four or five sample pieces on a simple website are: three blog posts and a landing page that demonstrate range, with a spec piece like what makes Franklin Barbecue distinct breaking the chicken-and-egg loop. The cold pitch is then four or five sentences — brief intro, one genuine detail that proves you actually consumed something they made, a link to the portfolio, and a short ask, where one writer landed a callback the same day after citing a specific moment from an agency's podcast. Beyond marketplaces, LinkedIn comments, niche communities and considered direct messages work when they are human rather than showy; $30 to $40 per hour is seen as a defensible Upwork starting band if the pitch and portfolio are tailored, project pricing follows scope capture via a notetaker like Granola, and the compounding picture is monthly posts at $800 to $1,200 plus college essay packages at $90 per hour totaling roughly $6,000 to $7,000 for a non-technical generalist who mixes product time, community listening and restrained model routing.