Stew Fortier is building “Google Docs from the future.” The co-founder and CEO of Type joins Jeremy Rivera on The Unscripted SEO Interview Podcast for a conversation that starts with model selection and ends somewhere considerably darker — the point at which AI-written content becomes the training data for the next model.
Along the way: why Claude and GPT get used for different jobs inside the same product, the prompt words that reliably pull better writing out of a model, and one concrete SEO tactic Fortier says changed the quality of his own keyword strategy work overnight.
Two models, two jobs
Type lets you switch between Claude and GPT, and the reason isn’t hedging — it’s that they fail differently.
“Claude can have these really great strengths with style and warmth and sounds a little bit more human and inviting. So in writing where that’s really important, we think that Claude is the right tool for the job… for harder writing tasks, where maybe there’s more complicated reasoning, or you’re writing something more technical in nature, something that depends on really airtight logic, the GPT models tend to be a little bit stronger.” — Stew Fortier
He’s also candid about why other models didn’t make the cut, and it has nothing to do with leaderboards: “Llama 3 will come out and it appears to be beating all these benchmarks. But when you actually go to use it, it just flubs on the simplest things… even though it cleared some machine learning benchmarks, the user experience is actually kind of subpar.”
Horizontal by design (and the surprise power users)
The bet behind Type was that writing tools are naturally horizontal — “both a poet and a lawyer can use Google Docs” — and that language models are horizontal too, equally comfortable with fiction and with processing a podcast transcript. Build the interface right and the same product serves podcasters, novelists, and lawyers.
That thesis proved out, with one segment Fortier did not forecast: “The surprising one that has insane usage is people writing extremely steamy romance novels… these are like the power users of the power users.”
The most forgetful co-worker ever
Asked for the roadmap, Fortier names the limitation honestly. “It’s like having the most forgetful co-worker ever. It doesn’t retain a lot of knowledge about you… it doesn’t know your style as well as it could.” The work in front of them is letting writers bring in outside knowledge, reuse it inside Type, and deepen the tool’s grip on individual style and preference.
Jeremy’s wish list is the SaaS operator’s version of the same thing: digest all the knowledge docs and support tickets, then use that corpus to shortcut blog content, ideate off the biggest support problems, and arm the sales team — connecting inputs and outputs across one lexicon of company information.
Where Type won’t compete — and where it will
On SEO specifically, Fortier draws a clear line rather than promising everything.
“There’s a world in which Type actually doesn’t have the best ideation or SEO strategy experience… but then when you’re ready to actually go create that content, Type is actually gonna be the superior authoring experience.” — Stew Fortier
The pieces that do belong in an authoring tool, he says, are the ones you need while writing: citing high-quality external sources and finding good internal links to reference. On connecting to a live web index — the capability Perplexity, Copilot, and ChatGPT search normalized — his read at the time of recording was that the problem had matured enough to adopt rather than pioneer: “my spidey sense is like, this is a this-year type of time range.”
Dogfooding, the honest way
The story of how Type’s own blog gets written is a small masterclass in inbound. An SEO writer discovered the product on their own, used it for client work, then pitched Type on improving the company’s own blog. They got hired. “So we’re dogfooding it in two ways… it’s a customer who’s now writing a lot of this content, [and] they’re using Type in their process.” Before that, Fortier wrote most of it himself — which, he notes, is how you find the gaps: “most of what you see is what’s yet to build.”
The qualifier-word trick
If you understand that a model is predicting the most likely next tokens, you understand why default AI writing feels bland — and how to steer out of it. Fortier’s fix is embarrassingly small and works better than it should.
“These small qualifier words — write this and include a surprising statistic instead of saying just include a statistic. Or give me a counterintuitive way that I could frame this idea. Or make this more interesting — literally you can just say that… I’ve been quite pleased with how much impact that can have on the overall output.” — Stew Fortier
The mechanism is the same one that makes rare vocabulary useful. When Jeremy throws in the word lugubrious, Fortier immediately clocks why it works: “that’s somewhere in the training data on these models. It’s probably not seen a lot. So if you use it, it’ll probably bring up these less likely, slightly more surprising responses.”
Personas vs. personalization
Jeremy floats the idea of named writing personas — a “Jeffrey” who never says “in conclusion” and never writes a throat-clearing intro. Fortier sees two roads: explicit modes (Hemingway mode, SEO best-practices mode), which plenty of products have shipped, or something harder and better — feed the model enough examples of what you consider good writing that it stops needing the instruction at all.
“It takes your prompting burden down quite a bit… it almost feels like it has a spidey sense because it’s seen enough examples of how you like to do this.” Jeremy’s summary of the distinction: personalization over time versus specialization.
The ouroboros problem
The philosophical turn: the last reliably human corpus on the internet is arguably pre-2012, and more of the marketing web every year is machine-generated or machine-assisted. What happens when models train on model output?
Fortier reframes it as a race rather than a doom loop.
“We’re currently in a race of sorts between the algorithms that blindly consume everything on the internet and learn from them, and the algorithms that have some ability to understand the underlying quality of that content and determine its usefulness… We have to get both things right.” — Stew Fortier
His analogy is the social feed: a lot of garbage never reaches you because a ranking system already judged it too uninteresting to surface. More AI content in the training data isn’t inherently fatal — but only if the judgment systems deciding what deserves attention improve at the same rate.
Hallucination and the silver-tongued fox
Fortier’s biggest worry about user behavior isn’t capability, it’s credulity.
“I notice this when I’m doing research or working with LLMs on something that I feel like I know pretty well — how often its advice or guidance or very plausible sounding suggestions are actually mediocre at best, but dead wrong at worst. And I think people are just trusting these things, and I do it too on stuff where I don’t know.” — Stew Fortier
Jeremy calls it “my father, the silver-tongued fox effect” — an answer delivered with total confidence, complete with a Mongol invasion of Serbia that never happened, which you only discover was wrong a year later in a history class. Both land on the same recommendation for anyone using these tools professionally: understand that this is prediction, not knowledge, and take the output with a grain of salt.
Fortier’s optimist case is the self-driving-car argument: if you can get a model marginally more factual than the average human, or the average subject matter expert, you want more of them on the road even knowing there will be accidents. “I think it’s an unbelievably hard problem. But if we can crack it, that could be great.”
Who loses work, and who gets hired back
On the employment question, Fortier is refreshingly unwilling to pretend he knows — but he reports a pattern he keeps hearing from other founders.
“Six months ago: hey, I just let go of my assistant in the Philippines and now I’m just using ChatGPT or Type. More recently: hey, I’m looking for a writer I can hire who’s using AI to be more productive because I need someone to manage all the AI workflows for me. And they’re essentially hiring back that person — but now they’re levered up with AI tools.” — Stew Fortier
The structural point underneath it: “if it’s now easier for everyone in the market to do X, then you’re going to get pushed to do Y.” The frontier moves; the work relocates to it. Jeremy’s addition is the credential angle — if the tooling lowers the barrier, the four-year marketing degree stops being the gate, which opens the field to people with drive and judgment who couldn’t buy the credential. Fortier extends it to founders: someone who can see a hole in the market but can’t hire an engineer now gets a few notches further on their own.
The one thing to do today: feed the context window
Asked for a single actionable takeaway for working SEOs, Fortier gives the most useful answer in the episode — and it costs nothing.
“These tasks that many of these language models fail at today… it often fails not because the model isn’t smart enough. It often fails because it does not have the context it needs, which you have in your head maybe, or you have sitting in your inbox, but it has not been exposed to.” — Stew Fortier
His own test: doing keyword strategy work for Type, he pasted in competitor sites, the actual SERP results for the target keywords, and current Search Console data. “I was kind of shocked at how much that impacted the quality of advice… It went from a borderline useless conversation to getting a couple of insights that I was actually willing to act on.”
The soundbite version, as Jeremy puts it: run your context window. In Type, the generate-draft feature accepts roughly 100,000 words across attachments, with documents you’re editing capped around 10,000 words.
Connect with Stew
Stew Fortier is co-founder and CEO of Type, an AI writing product built for people who write for a living — in-house, agency, or freelance. He posts on X as @stewfortier.
Listen to the full episode on Castos. For more conversations like this, explore the Unscripted SEO topical hubs or try the keyword-research tool at SEO Arcade.
