Every brand talks to a dozen audiences. Its references only work on one.
Vernacular is the pre-flight check for culture: it catches the shows, artists, brands and places in your copy that your audience won't get, and swaps them for ones they love, verified on Qloo's taste graph. Grammarly made everyone's writing correct. Vernacular makes it land.
Writers reach for their own culture, not their audience's.
A 45-year-old copywriter says “like a Friday at Olive Garden”. A 24-year-old reads it and scrolls on. Translation tools fix language, not culture, and generic AI copy makes it worse: models reach for the most famous reference, not the one this audience loves.
From our public benchmark: 8 pieces of copy × 5 audiences, every version graded the same way on Qloo. Method and every pair →
Four people who hit this problem every week.
The problem: One national script has to run on stations and feeds with very different listeners. Re-briefing a writer per audience takes days.
With Vernacular: Tour Mode scores the script for every audience at once; the agent localises each version; share cards go to the client.
See it live: A country-radio truck ad, rewritten for a Latin-pop station →The problem: Writes captions daily under time pressure; one dated reference and the comments turn on the brand.
With Vernacular: The extension checks the caption in the composer, a one-click swap fixes it, and predicted comments preview the reaction.
See it live: A TikTok skincare caption repurposed for 45–54 country fans →The problem: A billboard's references have to land with whoever drives past it, in that city.
With Vernacular: Local places are ranked among the city's well-known places, and the map shows which neighbourhoods the audience and each reference reach.
See it live: A Chicago sports-bar billboard for 30–34 hip-hop fans →The problem: Agents draft and publish copy at volume, and nothing checks whether it speaks to the audience.
With Vernacular: The MCP server lets any agent call check_copy before it publishes; the benchmark shows what that adds over the model alone.
See it live: Vernacular vs Claude alone, 40 pairs →One taste engine, everywhere copy gets written.
Around the check: hear the copy read aloud while a live reaction curve tracks each reference (the Dial Test), predict the comment section before posting, and see on a map which neighbourhoods a reference reaches for out-of-home campaigns.
The model writes. Qloo decides what lands.
A language model can rewrite a sentence, but it can't know what a specific audience actually loves. Qloo can: cross-domain taste affinity for music, film, TV, brands and places, by demographic, by interest and down to the neighbourhood. Vernacular puts every suggestion through that test, and shows its evidence.
Meaning is kept about as well as Claude alone keeps it (blind judge: 4.6/5 against 4.65/5). The moat compounds: every accepted or rejected swap is a labelled example of what works for whom, which no general model sees.
AI is writing the copy. Nobody is checking the culture.
Brands now publish more copy than ever, much of it drafted by models and soon published by agents. More output means more generic references and more audiences talked past. A taste check belongs in that pipeline the way spellcheck belongs in a word processor, and agents need it as a tool they can call.
Free in the browser, paid where the work happens.
| Plan (proposed) | For | What it unlocks |
|---|---|---|
| Free | Anyone writing | The extension's lint and a few full checks a month |
| Pro | Creators and marketers | Unlimited checks, saved audiences, Dial Test, predicted comments |
| Team | Agencies and brand teams | Shared brand audiences, approval flow, Tour Mode across markets |
| API | Agent and martech platforms | Usage-based MCP and REST access to check, translate and map |
The wedge is the free extension: it lives in every text box, so the product is discovered where the problem happens. Teams pay for shared audiences and approvals; platforms pay per check.
Where it goes from here.
- Your audiences, not presetsBuild audiences from a brand's own customer segments and Qloo's audience data.
- Every formatScripts, storyboards and video ads already work; next are slide decks, product pages and app-store listings at scale.
- Where teams writeGoogle Docs, Figma, Notion and CMS integrations alongside the extension.
- More languagesSpanish copy already works; the taste graph is global, so the same check extends to every market.