Launched Postessia today — an AI tool that writes LinkedIn posts in your actual voice (not generic AI-speak)
Launched Postessia today — an AI tool that writes LinkedIn posts in your actual voice (not generic AI-speak)
Hey founders ,
Live today: Postessia (postessia.in) — AI-powered LinkedIn content that sounds like you, not like every other AI post you've learned to scroll past.
Why I built this:
I kept seeing the same problem — founders and agency owners know consistent LinkedIn posting drives leads, but every AI writing tool produces the same flat, "unlock your potential 🚀" voice. You can spot AI content from a mile away now, and readers have started tuning it out.
So instead of "generate a post about X," Postessia analyzes how you actually write — your rhythm, your directness, the phrases you'd naturally use — and generates posts from that. Two-stage pipeline: first it builds a voice profile from your existing writing, then a planner→drafter→auditor loop checks every draft against that profile before it ships. If a draft doesn't sound like you, it gets rewritten before you ever see it.
What's live:
Free tier — 5 posts/month, no card needed
Solo — ₹599/mo
Founder — ₹1,499/month
What's not live yet: Agency/team plans — waitlist only for now. Didn't want to ship something half-built just to have more tiers on the pricing page.
I'm doing outreach manually right now (no automation, no bought lists) — genuinely trying to find people for whom this solves a real problem, not just chase signups. If you're an agency owner or founder who's given up on LinkedIn because you don't have time to write, I'd love your honest feedback — good or brutal.
postessia.in
Happy to answer anything about the build, the voice-matching approach, or why I skipped the "unlimited posts" pricing trap.
I can relate to that. It's rarely the writing itself it's the lack of original insight. Once content starts feeling generic, it's hard to stay engaged, regardless of how polished it is.
Spot on, mendy. You hit the nail on the head.A lot of AI writers focus entirely on making the prose look "polished" and grammatically perfect, but they completely strip out the author's unique perspective and raw insights in the process. It ends up reading like a textbook rather than a human sharing an experience.That’s exactly why we built the Auditor Loop to flag when the text starts losing that human edge and flattening into generic AI-speak.I’d love to drop you a beta key to test it out on your own drafts if you're open to it!
When you learn to pick up on the cues for AI-generated articles, it becomes incredibly easy to pick them out. I automatically lose all enthusiasm for engaging with the content.
If this works as described, I'm sure you'll do very well.
Spot on, Colin. That immediate 'turn off' when reading generic AI is exactly why we started building this. Word-banning list systems don't cut it anymore because readers spot the flat structural rhythm instantly.Would love to let you test it out on a draft or give you early access to see if it actually 'works as described' for your own workflow. Let me know if you're open to a beta key!
The planner-drafter-auditor loop makes more sense to me than banning a few cringe phrases and calling that voice. A lot of AI writing tools fail because they smooth everything into the same rhythm, even when the words are technically on-brand. I've seen the same thing on the dictation side: people don't mind cleanup, they mind when the tool starts rewriting them into a different person. That's part of why I built DictaFlow. The useful line is fixing the mess without sanding off the person's actual voice.
Thanks Ryan! 'Fixing the mess without sanding off the actual voice' is the absolute perfect way to phrase it. Traditional AI text optimization completely flattens the rhythm. DictaFlow sounds incredibly relevant here—capturing audio voice profile data is the hardest part. Let's definitely trade notes on how you tackle cleanup without losing the human edge
This resonates — I run a review blog (ai-tool-hunter.com) where every article is AI-generated, and "sounding like generic AI-speak" is the #1 thing I've had to actively fight against. I ended up hard-banning a list of words (revolutionize, game-changer, seamless, cutting-edge, etc.) directly in the system prompt, which helps but feels like a blunt instrument compared to what you're describing.
Curious about your planner→drafter→auditor loop specifically: what does the auditor actually check for when it flags a draft as "not sounding like you"? Is it comparing against banned phrases/patterns like my approach, or something more structural (sentence rhythm, paragraph length, how directly you make claims)? I've been meaning to build something similar for my own voice consistency and would love to know if word-banning is a dead end or if it's part of a bigger system.
Hey! Word-banning isn't a dead end, but it's definitely just step 1. The Auditor Loop looks past vocabulary. It maps structural DNA: typical paragraph pacing, how early you deliver a punchline, and your transition style. It catches when the AI tries to sound 'smart' instead of direct. I'd love to drop you early access to the auditor loop to test against your AI-generated articles. Let me know if you want a beta key!
Kind of ironic that we're both fighting the same "sounds like AI" problem from opposite sides honestly. The auditor loop checking drafts against your voice before you see them is a smart way to actually solve that instead of just promising it.
Spot on, Lily. The irony isn't lost on me—using an algorithm to make text sound less algorithmic! What 'side' of the problem are you tackling with your build? Would love to trade notes or give you early access to the loop to see if it holds up to your standards
Quick Day 3 Update: The global data is proving our core thesis right. Over the last 48 hours, this thread has pulled in traffic from the US, Canada, Portugal, and even guest post requests from Nigeria. It made me realize something fundamental about why generic AI slop is hitting a global wall. I just ran our internal launch data through Postessia to analyze the pattern. Here is the unedited reality of what we are fighting:
Generic AI content is flooding every channel startups use to reach people. It sounds competent. It ranks nowhere.
The problem isn't that AI-generated posts exist. It's that they all read the same. Same structure. Same transitions. Same three-word adjective stacks. Same engagement bait at the close.
LinkedIn's algorithm learned to spot it. Twitter learned to spot it. Your audience learned to spot it.
A founder I know spent six months publishing daily AI-written startup advice. Perfect grammar. Relevant keywords. Zero meaningful engagement. She switched to voice recordings of her actual thinking — rough, specific, occasional tangent — and her reach tripled in three weeks.
The cost of generic distribution is invisibility.
Here's what happens:
You sound like everyone else. Your differentiation disappears the moment you hit publish. Readers scroll past because they've seen the structure before.
Algorithms deprioritize it. Platforms now penalize content that matches known AI fingerprints. Generic is not just unnoticed — it's actively suppressed.
Trust erodes quietly. Audiences don't consciously think "this is AI." They feel it. Something is off. They move on.
Your unique insight gets buried. Even if your core idea is solid, the delivery drowns it. Generic wrapper kills specific insight.
The startups winning distribution right now aren't the ones with the most polished posts. They're the ones with voice.
Not storytelling. Not hacks. Voice. The particular way you see the problem. The specific number that surprised you. The detail nobody else would include.
That's what cuts through.
What is one insight about your startup that only you could articulate?
Why did you choose to target India rather than niche down in another way? Overall I like the concept and although others here are saying voice matching is table stakes I can't say I've seen it done well.
Thanks for digging in, Timothy! Appreciate you calling out the voice-matching gap. You're 100% right—a lot of tools claim they do it, but they usually just end up fixing grammar and spitting out polished corporate text that reads like an essay. To answer your question on targeting India as our initial niche:
The Attention Blueprint: The high-growth professional content market in India has a very unique psychological pattern. It relies heavily on conversational phrasing, sentence fragments, and raw vulnerability over hyper-polished "corporate-speak." If an AI engine can master this highly unpolished, voice-memo rhythm, standard Western corporate formatting actually becomes an easier problem to solve later.
The Velocity Testbed: India currently has one of the fastest-growing ecosystems of solo creators, freelancers, and agile B2B agencies globally. It is the perfect, high-volume testing ground for us to refine our Planner → Drafter → Auditor pipeline with real user friction. India is our launchpad to perfect the core engine under intense local constraints. We are already prepping our global rollout architecture (targeting a December release) because, like you noticed, the world is collectively exhausting its patience with generic AI slop.
Would love to know—what’s the biggest blocker you’ve noticed in other tools attempting voice matching?
Nowadays, AI dominates our daily lives, which greatly simplifies things. We just have to know how to choose the ones we really need, because more and more tools are being released every day. This is good for everyone, but it also increases our confusion about which one will completely and effectively solve our problem... Congratulations and good luck with what's to come.
Appreciate the good wishes, Joao!You hit the nail on the head. The massive explosion of new tools is exactly why buyers are facing extreme fatigue right now. Most software is just wrapping the same basic LLM API and outputting generic, robotic text. It complicates decisions instead of solving the core problem.Our ultimate goal with Postessia is to cut through that noise entirely. Instead of giving you another tool that makes you think about how to prompt, we want it to seamlessly adapt to your actual human voice right from the jump.Thanks for diving into the launch thread—stoked to have you following the build!
Congrats on shipping. The "actual voice, not generic AI-speak" angle is the whole game right now — "AI writes X" is commoditized, so the defensible part is the constraint you put on top, and the planner → drafter → auditor split is a smart way to enforce it. I'm building in a different space (turning books people own into summaries) and hit the same lesson: the feature isn't "AI summarizes", it's the constraint ("works with the book you actually own, not a catalog"). Question on the auditor stage — how does it judge "this sounds like me" vs just "this is grammatical"?
Scoring voice-match objectively feels like the hard part, and where it holds up or breaks as people post more.
It's rule-based, not a learned score. Voice profile has an explicit AI-tell audit 8 known patterns (triplet-adjective stacking, em-dash overuse, throat-clearing transitions, symmetrical sentence pairs, etc.) each one tagged per-user as authorized / flag-as-overused / banned, based on whether that writer naturally does it or not. Generator follows that over its own judgement, plus a hard rule: if 2+ banned patterns land in the same post, it has to rewrite before shipping. The "grammatical but not me" layer structure, sentence rhythm, paragraph density is matched separately (±15% char count, same paragraph skeleton, same sentence-length variation). Still not solving true semantic voice-match, that's the harder problem. Might write this up properly once I have more real output data.
Voice authenticity is the real moat here. How are you handling the "sounds like me but isn't" disconnect that kills adoption for most AI writing tools?
Two-stage split: Voice Analyst reads 1-3 of your real posts and extracts a structured profile tone, structure, sentence rhythm, vocabulary tics, and specifically which "AI-sounding" patterns you naturally use vs which read as fake if added. Ghostwriter then generates against that profile, not a generic prompt, and checks its own output against a banned-pattern list before shipping auto-rewrites if it stacks too many AI-tells. The "sounds like me but isn't" gap is exactly the thing this is built to close - still imperfect, but it's a real mechanism, not just a style instruction.
You're pointing at a real problem. People can smell "AI voice" immediately now, and most tools still optimize for volume instead of preserving the weird little phrasing that makes someone sound like themselves. One thing that helped me with DictaFlow was treating rough input as sacred for as long as possible. Capture the thought in the person's natural rhythm first, then clean it up, not rewrite it. If Postessia keeps that discipline, the product will feel a lot different from the generic post generators.
"Rough input as sacred" is the right instinct — most tools optimize the cleanup step and lose the person in it. Postessia runs two stages for that reason: extract the voice profile first, generate against it second. Still catching some AI-tell patterns slipping through, but that's the target. Stealing your line as a review checkpoint.
"Your tiers already think in stages, you just haven't tested them against one"
Good catch above on the ICP gap. I'd push one layer further. Free → Solo → ₹1,499 Founder is already a stage ladder, you just built it on instinct. The open question: is a 2-person Indian agency at pre-seed actually going to absorb ₹1,499/mo before they've landed 3 retainer clients, or does that tier only make sense once they're Seed-stage with 8-10 clients? That's the kind of thing that's hard to gut-check alone — worth mapping your tiers against actual company-stage spend patterns before you lock the agency tier pricing.
Honestly, fair challenge — I built the ladder on instinct (Free → Solo → Founder tracking typical agency growth), not on actual spend data yet. My guess is you're right that a pre-seed 2-person shop probably starts at Solo, not Founder, and grows into it once they've got retainer clients to justify workspace/reporting features. But that's a hypothesis, not something I've validated. Going to watch which tier early signups actually pick and revisit in a few weeks with real data instead of guessing further. Appreciate you pushing on it
Exactly. Here's the shortcut: list Postessia on SoftRankings (stage-fit tool directory). You get founder traffic pre-segmented by stage — Pre-seed separate from Seed separate from Series A. Watch which tier each cohort picks in week 1, not weeks. Saves guessing.
Here is a sample analytics page from Inssist on SoftRankings : https://softrankings.com/products/inssist/analytics
Appreciate this, will get Postessia listed there. Good instinct on watching cohort behavior by stage instead of guessing — that's basically what I'm doing internally too, just needed a cleaner way to track it. Thanks for digging in twice now.
Wow, you already launched on SoftRankings? Congratulations mate!
https://softrankings.com/products/postessia
Ran your landing page and one thing's worth flagging above all: your post pitches "founders and agency owners" broadly, but the page is built specifically for Indian agencies. ₹ pricing, "Desi Grind" tone, "Built for India." That's your sharpest wedge, but the two don't match. "AI writes LinkedIn posts in your voice" competes with 50 tools. "LinkedIn content for Indian agencies" competes with almost none. The India angle is the moat, lead with it everywhere, including here.
Second flag: your strongest features (client workspaces, approval queue, white-label reports) are all agency features, but locked to the priciest Founder tier. Your actual buyer pays the most to get the thing built for them. Worth rethinking.
Voice-matching is table stakes now. The Indian-agency angle is the real story. Who's the first agency that paid, and which tone did they pick?
Appreciate you actually digging into the page instead of just skimming the post — this is the kind of feedback that's useful.
On the positioning gap — you're right, and it's the sharper insight here. I built the product with Indian agencies as the real ICP, but the post copy defaulted to broader "founders and agency owners" language instead of leading with the actual wedge. That's a messaging gap, not a product gap, and it's an easy fix — headline and above-fold need to say "Indian agencies" explicitly instead of implying it through ₹ pricing and "Desi Grind" tone. Fixing that this week.
On the feature-tier point — this one I'll push back on slightly. Client workspaces, approval queue, and white-label reports are locked to Founder tier deliberately: an agency handling even 8-10 clients is looking at ₹2,999+/month value easily, so the tier is priced against agency-scale usage, not against a solo creator's willingness to pay. That said, those specific features are still in development, launching soon — so right now it's more of a roadmap commitment than a live mismatch. Once they ship, I'll be watching if agencies actually convert into Founder tier at that price point or if the tier needs adjusting. Real signal over assumption.
Good push on both fronts, first one's getting fixed now.