Freak Lab, the AI tools we build in house.
AI is only as good as the data you put in and the algorithm you give it to work with. So that is where we put our effort. Freak Lab is our everytool kit for marketing, built on decades of consumer research, datasets we pay for and 20 years of platform knowledge.
AI is only as good as its data and its algorithm.
Everyone has access to the same large language models. What changes the result is what you feed them and the rules you give them to follow. We focus on those two things, and we are prepared to pay for both.
We buy datasets to sharpen our tools, and we add our own. Decades of eye tracking and consumer behaviour research, licensed keyword and market data, 20 years of Amazon platform knowledge and the results of our campaigns in more than 100 countries.
- Paid keyword, market and attention datasets
- Our own test results, by category and market
- Client data kept inside that client's own group
Every tool runs on a framework we wrote. Our neural marketing framework for images, information hierarchies for listings, a moderation framework built from platform guidelines and a routing harness that picks the right model for each job.
- Frameworks written by our specialists
- The right model for each task, not one for everything
- A person reviews every output
Six tools, one lab.
Each tool started as a problem on a client account. Now they run behind everything we make.
Image Analyser
Our neural marketing framework predicts where a shopper's eye lands before they've had time to think. It was devised from decades of eye tracking and consumer behaviour data, and it scores every image on the five things that decide whether it gets noticed and understood.
Upload an image and say where it will be used. A saliency model trained on eye tracking data draws the heatmap and the eye path, and a language model writes the analysis against our framework, with a score out of 100 and the fixes that would raise it. The same image always gives the same map.
- Visual saliency, the instant grab
- The three second rule and cognitive load
- Trust anchors and gaze bias
- F and Z reading patterns
- Colour psychology and contrast
Freak Copy
Freak Copy writes Amazon titles, bullet points, descriptions and backend keywords from your product claims, your brand voice, live listing data and keyword data we license. It builds an information hierarchy first, so the words that matter most to shoppers lead.
For an existing listing, choose how far to go. Light for small amends, Normal for a Keyword Boost® that adds the most search terms without losing your voice, or Aggressive for a full rewrite. Across a range it covers core terms and emerging ones without products competing for the same searches.
- New listings or updates to live ones
- Keyword coverage shown before and after
- Brand voice set once, applied every time
Freak Reviews
Ratings of four stars and above change what shoppers buy, and one early review can sink a launch. Our review moderation process puts 20 years of Amazon platform knowledge into an AI flow that finds the reviews breaking the rules, so they can be reported and actioned.
It downloads every review and checks each one under three stars against a moderation framework we built from Amazon's Community Guidelines and industry review standards, then gives it a code. It drafts the report for each claim, ranks the ASINs where moderation would lift the rating most and flags patterns that point to bots or competitors for escalation.
- Every review under three stars checked
- Moderation codes and draft reports
- ASINs prioritised by rating and sales volume
Freak Content
AI production, used where it helps and checked where it matters. Our models are trained on rights cleared shoots, so a product can be placed in new scenes and formats without new photography, and every asset carries a record of the tools that touched it.
Before anything ships it runs through automated quality checks for spelling, grammar, claims and platform specs, in every language it's been transcreated into. Assets that need an AI label under the EU AI Act or a platform's rules are flagged, so nothing goes live without the right disclosure.
- Quality checks at scale, in every language
- AI labelling flagged automatically
- Impact assessments before AI is used at scale
Freak Creators
Freak Creators runs our influencer, seeding and gifting programmes. It classifies each creator's audience, records their vetting scores and follows them through all nine stages, from the first message to the post and the results.
Posts that tag the brand are picked up through the platforms' own data and checked for an #ad label before they count. Reminders go out on day seven and day twelve of the fourteen day window, and the best posts are flagged for a rights request and a paid partnership test.
- Audience checks before a creator is approved
- One tracker for every market and client
- Posts found and checked automatically
Freak Games
Our games engine builds branded quizzes, games and learning modules from templates in any language and records every answer. Points, levels and leaderboards keep people playing, and the data shows what they learned and where they dropped off.
The same thinking is in RNIB's Learning Portal, which we designed and built, where four modules teach sight loss awareness through film, quizzes and a sight simulator, with points, an anonymised leaderboard and shareable certificates.
- Games and quizzes from templates
- Points, levels and leaderboards
- Live data on every player and question
From a client problem to a tool every client uses.
- Problem
A task on a client account that's slow, manual or guesswork.
- Data
The datasets that would answer it, licensed, owned or the client's.
- Framework
The rules our specialists would follow, written down.
- Models
The right model for each step, routed through our harness.
- Review
People check every output, and their edits teach the tool.
- Measure
Results tracked against the old way, before it goes wider.
The winners won't be the brands that use AI the most.
They'll be the ones who know where to use it and where not to. Our own impact assessments found that when real content carried a made with AI label, social engagement fell by 15 - 20% and ecommerce conversion by 13 - 16%, and premium brands were hit hardest.
So we use AI freely behind the scenes and carefully in front of shoppers, and we test before we scale. Stuart Willison, our co-founder and CEO, set out the thinking in an interview with LBB.
Read our view on AI and brand value| Where | How we use AI |
|---|---|
| Analysis and research | Freely, with people checking the conclusions |
| Ideation and creative development | Freely, as a starting point for our creatives |
| Quality checks and transcreation | At scale, with native speakers on sign off |
| Ecommerce content | Carefully, after an impact assessment |
| Brand campaigns and social | Only by agreement, labelled where required |
Where Freak Lab has already paid back.
From the Resource Centre.
How we score Amazon images before shoppers ever see them.
The five measures behind our neural marketing framework, what each one tests and how we use them to fix Amazon images before they go live.
Amazon review moderation, and how to protect your star rating.
Why four stars matters, how we find reviews that break Amazon's guidelines and the six step process we run for clients.
Brand value, not compliance, is the biggest AI risk.
What our impact assessments found when real content carried a made with AI label, the rules now in force and six steps to use AI without hurting the brand.
Questions about our AI.
What is Freak Lab?
Freak Lab is the set of AI tools Freak builds and runs in house, our everytool kit for marketing. It covers image analysis, Amazon copy, review moderation, content production, creator seeding and branded games, and it sits behind the work we do for clients.
How does Freak use AI in marketing?
We use AI where it makes the work better or faster without hurting the brand. That means analysing images and copy, moderating reviews, quality checks, transcreation and building custom tools. We are careful about AI in high visibility brand work, because our research shows labelled AI content can lower engagement and conversion.
What is the Freak neural marketing framework?
It is our method for predicting how shoppers look at an image. It was devised from decades of eye tracking and consumer behaviour research, and it scores five things. Visual saliency, the three second rule, trust anchors, reading patterns and colour psychology.
Why does Freak pay for data?
Because AI is only as good as the data you put in. Licensed keyword, market and eye tracking data, plus our own campaign results, give our tools a better starting point than a general model working from the open web.
Can clients use Freak Lab themselves?
Yes. Access runs through groups set up by client and country, so each team sees only its own brands, analyses and history.
Do you label AI generated content?
Yes, where the rules require it. We follow the EU AI Act and each platform's rules, run a standard quality check on every AI assisted asset and keep a record of which tools touched it.
Will AI replace your creative team?
No. Freak Lab does the analysis and the repetitive work, so our creatives spend their time on ideas. Every output is reviewed by a person before it reaches a client or a shopper.
Want to see Freak Lab on your brand?
Send us a product page or a set of images, and we'll run them through the tools for you.