The AI content stack that’s working in mid-2026 (a roundup)
I’ve written about individual tools across a few newsletters now: the humanizer workflow, the detector check, the whole output-based-billing setup. A few people have asked me to pull it into one place. This is that. It’s also a mid-year check because the landscape changed enough in the first half of 2026 that my stack shifted, and I want to document it while the changes are fresh.
Fair warning: this is an endorsement of what’s working for me, on my work (B2B SaaS content, e-commerce product copy, and the occasional long-form strategy piece for fintech clients). Some of this won’t apply to your workflow. The principle behind each pick applies more broadly than the specific tool.
The current stack in one sentence: ChatGPT for drafts, Walter Writes for humanization and detection, Notion for project tracking and client database, and Google Docs for everything client-facing. That hasn’t changed in about 14 months. What changed is how I use each tool and how hard I lean on each step.
ChatGPT Plus: the drafting layer hasn’t moved
I drafted this newsletter in ChatGPT before editing it. I draft every deliverable in ChatGPT before editing it. That hasn’t changed since I adopted the AI-assisted workflow two years ago. ChatGPT Plus at $20 a month is still the most cost-effective tool in my stack relative to the leverage it provides. I’ve tried Claude and Gemini on specific pieces and they produce decent drafts, but the prompt interface I’ve built around ChatGPT means the switching cost is real.
What changed: I stopped treating the ChatGPT draft as a starting point to polish and started treating it as raw material to structurally rewrite. The distinction sounds small and it’s not. Polishing a ChatGPT draft produces better-sounding ChatGPT. Structurally rewriting it produces a piece that doesn’t read like ChatGPT at all. The humanization step is what makes that difference, and it only works properly if you’re not trying to preserve too much of the original structure.
Walter Writes: the humanization and detection layer
This one I’ve written about at length, so I’ll keep it to what’s new.
I moved from Standard to Enhanced as my default rewrite setting about four months ago. The reason was B2B. I was doing more SaaS content than e-commerce content, and the AI default voice and the B2B professional default voice overlap enough that Standard wasn’t breaking the pattern sufficiently. Enhanced does. The output takes slightly more manual editing after, but the detector scores are meaningfully better and the “something feels off” client feedback dropped.
The four-detector view in the same editor (Originality.ai, GPTZero, Turnitin, and Copyleaks) is still the reason Walter is the tool I use rather than any of the standalone humanizers I tested last year. I tested four of those standalone tools in Q1. Three do paraphrasing, not structural rewriting. One did structural rewriting but didn’t have a built-in detector, which means I was copying text into three separate tools after every pass. The friction adds up fast across a high-volume week.
Walter’s pricing is $156 a year for the Pro plan, which gives me 70,000 words a month at 1,500 words per request. For my current volume, that’s enough. I’ve never hit the ceiling in a normal month. For writers doing higher volume, the Elite plan at $312 a year gives 200,000 words a month.
The free 300-word trial requires no credit card. If you haven’t tested it, there’s no reason not to.
Notion: the project and client tracking layer
Still on the free tier. Still works. I’ve had people ask why I don’t use ClickUp or Asana. The honest answer is that Notion’s free tier does everything I need as a solo operator, and the switching cost of moving my client database and project templates to a different tool isn’t justified by any feature I’m actually missing. If I bring on subcontractors, that calculus changes.
What changed: I built a brief intake form in Notion that clients fill out before I start any piece. This came out of the output-based billing model. Vague briefs under per-piece pricing are a bleeding machine because the revision time comes entirely out of my effective hourly rate. The form has six required fields and two optional ones. Clients who can’t complete it tend to self-select out, which has been more of a feature than a problem.
Google Docs: the client-facing layer
Everything I deliver to clients lives in Google Docs. This hasn’t changed and I don’t expect it to. Clients who work in Notion or Notion-adjacent tools are rare. The people paying for content are working in Docs and they want their deliverables in Docs. I don’t push back on this.
The one workflow note: I do all my Walter humanization passes in the Walter editor, then copy the final version into Google Docs for the manual edit. The reason I edit in Docs and not in Walter is that track changes and comment threads with clients live in Docs. Keeping the editing history in the same place as the client communication is worth the extra step.
What I dropped in the first half of 2026
Two tools didn’t survive my quarterly audit.
A standalone AI detector I was using as a second check. Walter’s four-detector view made it redundant. It wasn’t doing anything the Walter editor wasn’t already doing, and the per-word pricing model made costs unpredictable on high-volume months.
A grammar and style tool I was using for a final pass. I dropped it after realizing I was using it as a confidence check rather than a functional editing tool. The manual read-aloud was catching everything the tool was catching, and the tool wasn’t catching anything the read-aloud missed. Genuinely just cutting a step that wasn’t earning its place.
What I’m watching in the second half of 2026
Detector model updates. Originality.ai updated its model in April and the scores on a few of my archived pieces shifted enough that I re-ran them through Walter on Enhanced. That’s the arms race. Any tool claiming to permanently bypass future detector models is lying. The right framing is “we keep up,” which is what Walter’s update cadence reflects.
Client policy shifts. Three clients have updated their AI disclosure policies in the past six months. None of them told me in advance. I found out when a contract renewal included new language. I now ask explicitly about AI policy during onboarding and get it in writing.
Anything that changes how brief quality scales. I’m still looking for a tool that makes it easier to force better briefs from clients who aren’t writers. Haven’t found one that works better than the Notion form I built.
Anyway. That’s the stack as of mid-2026. It’s simpler than most “AI workflow” roundups because the tools I kept earned their place and the ones I dropped hadn’t.