The claim, stated plainly
Ask your market anything. Theia reads what the internet actually says about it — every relevant source (articles, video, forums, reviews, and the AI answers themselves), every competitor, every language the market speaks — and answers with evidence you can check. Not a survey of a thousand people; the recorded voice of the whole market.
That is a big claim, and the right response to a big claim is examples. Below: seven real questions, from seven of the twelve live markets in our store — 816,461 evidenced observations and counting — each answered by the same engine, each answer carrying its numbers. Where the market speaks German or Welsh, so does the evidence.
1. "Which weight-loss treatment do UK consumers actually want?" (demand)
Mounjaro — and it isn't close. 1.87M UK searches a month, five times Wegovy (370K), eight times Ozempic (234K); Mounjaro alone out-demands every generic term in the category. The market is already pre-selling itself the next molecule: retatrutide draws ~90K monthly searches before launch. Evidence: 1,773 keywords priced on an 87-month measured demand series. → The weight-loss market study
2. "Who do AI answers trust about weight loss?" (AI visibility)
Not who Google trusts. NHS.uk holds 15.4% of classic Search visibility but just 1.5% of AI-answer citations; NICE — with effectively zero Search presence — takes 5.4% of citations. The AI layer re-ranks the market, and we measure both layers on the same queries. Evidence: CTR-weighted Search share vs measured AI-citation share, per domain. → The AI reshuffle, measured
3. "Is the US cannabis market just a bigger UK?" (two-country comparison)
No — it's a different market wearing the same plant. UK demand splits medical 49% / wellness 47% / recreational 4%; the US splits wellness 53% / recreational 26% / medical 21%. Britain's lead use-case is epilepsy (the licensed-medicine effect); America's is sleep, at 1.1M searches a month. Evidence: 155,012 voices + 87-month demand series, both countries, one taxonomy. → Cannabis: UK vs US
4. "What do German listeners say premium audio brands get wrong?" (perception, in German)
The engine reads the German audio press and forums natively — no translation pipeline, no English-only sample. What earns German praise is concrete: "Es gibt hier ein Maß an Klarheit und Tiefe, das wir auf diesem Niveau noch nicht gehört haben" — clarity and depth, at a price level, benchmarked. The German market rewards engineering specificity over lifestyle marketing. Evidence: 36,669 observations from the German audio market, read in German. → The sound market study
5. "How do Welsh-medium teachers experience their exam board?" (a market that speaks Welsh)
The Welsh-language education conversation happens in Welsh — "Rydyn ni wedi gweithio gyda'r sector i ddatblygu adnoddau Sgiliau Hanfodol Cymraeg" — and the engine reads it alongside the English. The finding: the board leads where it has statutory duty (Wales, access arrangements) yet the Welsh curriculum-reform narrative largely happens without it — 3 mentions in 2,745 reform observations. Evidence: 74,139 observations across the UK education conversation, English and Welsh, one analysis.
6. "Where does a loved regional bank lose its customers?" (experience diagnosis)
In the app. Customers score the human service +0.71 and account setup up to +0.90 (on −1 to +1) — and the mobile app −0.52. A brand whose people are loved and whose digital experience is the drag: the single highest-leverage fix, found by reading what customers wrote, not by asking them to fill in a survey. Evidence: 96,630 observations across the UK building-society market. → The building society study
7. "Who does AI cite about market research itself?" (the reflexive test)
We point the engine at our own category, by the same rules: the AI answers cite NielsenIQ, Kantar, Qualtrics, Mintel, the MRS — and we track our own citation share with the same instrument, in public. Evidence: the live category measurement behind "AI answers have a shortlist".
The hard numbers behind every answer
Narrative is cheap. Every Theia answer stands on measured dimensions — the same five, in every market:
| Measure | What it looks like, measured |
|---|---|
| Demand trends | Mounjaro: 0 → 1.87M monthly UK searches since its 2022 entry, on an 87-month measured series. US cannabis: 58M → 79M searches/year to 2023, then plateau — the growth era's end, visible in the data. |
| Video reach | Counted on evidence-producing videos only — never raw discovery sums (which run 7–100× higher on noise). UK weight loss: 422 relevant videos, 125.7M views — and the reach ranking inverts the search ranking: Ozempic leads video reach at 40.6M views vs Mounjaro's 24.3M, while Mounjaro leads search 8-to-1. Fame and purchase intent are different metrics; we measure both. |
| Sentiment, scored | Weight regain after stopping: Mounjaro +0.45 vs the semaglutide family −0.34 (−1 to +1). A building society's staff: +0.71; its app: −0.52. Positioning read straight off the axis. |
| Where the conversation happens | Weight loss: 30% of voices on Reddit. Cannabis: 3%. Education: 73% open web, 11% YouTube. Same instrument, different market shapes — channel strategy falls out of the split. |
| Who is talking | Weight loss: patients 53% vs clinicians 47% of role-claimed voices — two conversations, measurably different. Education: professionals 93%, learners under 1% — the student voice is largely absent from the open conversation, which is itself a strategic finding. |
Each observation carries all of it at once — source, platform, language, sentiment, speaker role, emotional register — which is why a question like "what do clinicians (not patients) say about side effects?" is a filter, not a new project.
What comes out the other side
Answering questions is the engine. What you walk away with is the point — every study ships as a working set of outcomes, not a PDF:
The analysis — the full market read: the deck that tells the story, the strategic reports behind every slide, the maps, and the data workbooks with every number's provenance. See the outputs →
A dedicated chatbot — with every study. Not a demo: a scoped analyst trained on your study's corpus, answering only from its evidence, quotes attached. Your team asks the questions we didn't think to — months after delivery. The two above are live right now: weight-loss · cannabis. No one else in the category hands you the study and the analyst.
Product positioning — where each product wins and loses, on the axes the market actually decides on: the perception leaderboards, the head-to-head matchups, the whitespace nobody owns. (The weight-loss study's "regain moat" is a positioning call, read straight from 79,000 voices.)
Marketing strategy — the four pillars turned into the strategy read: which demand to chase, which stage of the journey to own, where the AI-answer layer re-ranks your category, and what to do in what order.
Content, benchmarked and measured — the part most agencies won't submit to: content built from the evidence, scored against the market before it ships, gated on groundedness (every claim traces to a source), and re-measured after — rankings, citations, coverage. If it can't be measured moving, we don't call it done.
Why this works — and why a sample can't
Every answer above comes from the same architecture: exhaustive collection (the whole market's public voice, not a panel), cross-language harmonisation (German, Welsh, French, Italian and English observations land in one comparable analysis), measured demand (real monthly series, not a tool's index), and groundedness — every claim traces to quoted evidence. The chatbots exist because the evidence does.
Ask it something we haven't answered here: the weight-loss chatbot · the cannabis chatbot — or bring us your market.