Help & FAQ

Everything BuyerPanel does, explained.

One loop, five steps: create ads, prove them on a synthetic buyer panel before you spend, evolve the winners, ship to Meta, and learn from real results.
The buyer panel Creating ads Testing & evolving Shipping to Meta Learning from results FAQ
THE SCIENCE

The synthetic buyer panel

Every ad in BuyerPanel is judged before it costs you anything - here's why that judgment is worth trusting.

SSR · SEMANTIC SIMILARITY RATING

Survey-grade purchase-intent scores, in minutes, for $0

Each synthetic buyer reacts to your ad in words - the way a real person would in a focus group. Those free-text reactions are converted into a 1–5 purchase-intent score using SSR, a method validated on 57 real product surveys with 9,300+ human responses. The result: AI panels reproduce human purchase-intent ratings at ~90% of test–retest reliability - statistically, the panel agrees with human panels about as well as human panels agree with themselves when the same survey is run twice.

You don't just get a number. Every ad shows its score distribution (consensus vs polarizing) and every persona's reaction in their own words - the why behind the score.

AUDIENCE PANEL

You decide who judges

Pick one of 16 ready-made audiences (ages 16–80, US & EU - teen gamers, new parents, small business owners, luxury shoppers…), upload your Google Ads / Analytics demographics export, or simply describe your buyers in a paragraph. BuyerPanel builds a panel of synthetic personas from it, and every scoring run - everywhere in the app - uses those people. Hover any persona card to remove individuals; without a custom panel, one is auto-generated from your business description each run.

STEP 1

Creating ads - no designer needed

From a URL to a board of ranked, ready-to-render ad concepts.

ONE-URL ONBOARDING

Paste your website, get a filled-in brief

Drop your website / Shopify / App Store URL into the import bar (or the homepage hero). BuyerPanel reads the page - including bot-protected sites, via a browser-grade fetch chain - and fills in your brand brief, target customer, audience description and reference angles. With an Anthropic key you get the full extracted brief; without one it imports title and description and tells you exactly that.

REFERENCES & COMPETITOR INTEL

Steal like a strategist

Three ways to point generation at what already works: paste reference ad text, pull competitors' live ads from the official Meta Ad Library (free token), or drop up to 12 winning ad visuals - the vision model reads each creative and folds its hook, layout and angle into the brief.

GENERATE & RANK

Concepts land pre-tested

Hit Generate & rank ads: BuyerPanel writes distinct concepts - different angles (problem-led, founder POV, social proof, mechanism…), awareness stages, headlines and primary text - and the panel scores every one instantly. They arrive as a ranked scoreboard with a TOP PICK, score distributions, and each persona's reasoning.

PER-AD TOOLKIT

From concept to finished creative

STEP 2

Testing & evolving

Dozens of ads tested, $0 spent - winners mutate, losers die.

⚡ EVOLUTION MODE

The auto-loop

Choose 2–4 generations and press Evolve. Each round: generate → the panel judges → the top ads mutate (sharper hook, added proof, identity angle, urgency) while losers join the kill list → repeat. You get lineage boards per generation, an improvement curve, and a headline number like “gen 3 beat gen 1 by +30% intent”.

RANK MY ADS

Already have ads? Test them first

Paste existing ads one per line (Headline | Primary text), drop a CSV, or drop finished image ads (PNG/JPG, multiple at once - the panel sees the actual creative). Know which one to run before you spend.

LANDING PAGES

Test the page, not just the ad

The click is only half the funnel. Paste up to 5 landing-page URLs - your variants, or a competitor's page - and the same buyer panel reads each one as people who just clicked your ad: a purchase-intent score per page, the panel's reactions, and the top conversion fixes. Compare before you spend on traffic.

STEP 3

Shipping to Meta - always paused

One click puts the ad in your Ads Manager. Nothing ever spends without you.

SEND TO META

Paused drafts in your own ad account

Every ranked ad has a Send to Meta action: BuyerPanel creates the campaign, ad set, creative and ad in your own Ads Manager - all with status PAUSED. You review and press go there. Setup is a one-time ~10 minutes with your own Meta app in developer mode (no Meta App Review needed, because it's your app managing your account): create an app at developers.facebook.com → add the Marketing API → generate a token with ads_management → paste token, ad account ID and Page ID into Settings. The in-app guide walks you through it.

STEP 4

Learning from real results

This is the flywheel: the panel gets measurably better at predicting your customers.

RESULTS LOOP

Drop your results CSV, get the next round

After your winners run, export the ad-level results from Ads Manager and drop the CSV into the Results loop. Winners become references, losers go on the kill list, and the next round generates automatically. Keep the ad names BuyerPanel gave them - that's the join that makes it work.

CALIBRATION

SSR vs your real ROAS, tracked per project

Every upload computes the correlation between the panel's predictions and your real results (Spearman's rho - ≥0.6 is strong). It sharpens with every round: compounding, brand-specific accuracy nobody can copy.

FORECAST & LIBRARY

Project forward, look back

Forecast takes any dated metrics export (Meta / Google / TikTok day breakdown) and projects each campaign forward with uncertainty bands. Library keeps every run reopenable - boards, generations, images, and the panel's words.

QUESTIONS

FAQ

Short answers to the things everyone asks.

How accurate is the ad testing, really?
The method (SSR) was validated by PyMC Labs in 2025 on 57 real product surveys with 9,300+ human responses: AI panels reproduced human purchase-intent ratings at ~90% of human test–retest reliability - meaning the panel agrees with human panels about as well as human panels agree with themselves. It's a pre-test, not an oracle: use it to kill weak ads and rank strong ones before spending, then let the Results loop calibrate predictions against your real ROAS.
Do I need API keys? Which ones?
BuyerPanel is bring-your-own-key, so you pay providers directly at cost. Anthropic unlocks real ad generation and persona reactions; Voyage or OpenAI unlocks real SSR scoring (those two together switch the badge to LIVE); fal.ai and/or OpenAI unlock image rendering; HeyGen unlocks real avatar videos; free Meta tokens unlock competitor intel and publishing. Keys are added in ⚙ Settings, take effect instantly, and are masked after saving.
What happens before I add any keys?
Everything runs in demo mode: the full flow is clickable with clearly-labelled placeholder data, so you can learn the product without configuring anything. BuyerPanel never passes fake output off as real - demo copy says it's demo copy, and features that can't be faked honestly (like avatar video) tell you which key they need instead.
Can BuyerPanel spend my ad budget?
No. Publishing creates everything in your Ads Manager with status PAUSED - campaign, ad set and ad. You review and press go inside Meta. There is no path in the product that starts spend on your behalf.
Why do I connect Meta with my own app instead of a "Login with Facebook" button?
A central login button requires Meta's App Review (weeks of process). Your own app in developer mode can always manage your own ad account with no review at all - so you're publishing in ~10 minutes instead. The default token lasts ~60 days (re-paste when it expires); a Business Manager system user token never expires.
Can I test ads I already have?
Yes - that's the Rank my ads screen. Paste them one per line, drop a CSV, or drop finished image ads (multiple PNG/JPGs at once). The panel judges the actual creative, scores each ad, and explains every score.
How do the UGC videos work? Are they real?
Real. BuyerPanel writes a panel-scored UGC script, then renders it through HeyGen with your key: you pick from hundreds of stock avatars and multilingual voices, and get a lip-synced 9:16 video (typically 1–5 minutes to render). There are no fake video previews - without a HeyGen key the app says exactly that.
Are my API keys and data safe?
Keys are stored per-account, never shown again after saving (fields display ********), and are only used server-side to call the providers you configured. Your projects, panels, runs and results are scoped to your account.
What do credits do?
Each action meters a small credit cost (generate 5, rank 2, image render 5, avatar video 25…). Your balance lives in the sidebar and the Usage panel in Settings shows recent activity and a top-up link when billing is enabled.
Which ad platforms are supported?
Publishing (paused drafts) targets Meta today - competitor intel also comes from the Meta Ad Library. The Results loop and Forecast accept exports from Meta, Google and TikTok, and the export pack gives you creative + copy to run anywhere.
How is this different from Superscale, AdCreative and the other AI ad tools?
Two things. Proof before spend: others generate creatives; BuyerPanel generates them and pre-tests every one on a research-validated buyer panel, with the why behind every score. The calibration flywheel: your real results feed back in, so predictions keep getting better for your specific brand - an advantage that compounds and can't be copied.

Meet the panel before your customers do.

Your first run is free - and the demo works with zero setup.

Get started free