#2 retailer
most-cited in AI fragrance answers (faces.ae)
Faces already shows up in AI fragrance answers. GEO measures it, then grows it.
A growing share of beauty discovery now happens inside an AI engine (ChatGPT, Google AI Overviews), not on a results page. Faces already appears there: in fragrance answers it is the #2 retailer cited, strongest on Google's surfaces. But that presence is unmeasured and unmanaged. This POC, run with Profound, establishes a proper baseline across Perfumes and Skincare, sizes the gap, and uses AI agents to target a ~75% visibility uplift.
- 1
Baseline
Establish the starting point
How often Faces appears across AI engines, for Perfumes + Skincare.
- 2
Size
Find the gaps
Which categories and prompts have the most room to move.
- 3
Optimize
Run the agents
Profound Agents win the answers we're currently missing.
- 4
Prove
Measure the uplift
Visibility gain vs a holdout, plus the AI-driven traffic it sends.
Where we are: an early read is in (fragrance only): Faces is the #2 retailer cited, ~77% positive, strongest on Google. The full baseline across Perfumes + Skincare takes a few weeks to stand up. Kickoff is gated on legal + infosec review; agent build can begin in the demo environment in parallel.
Where Faces stands today
Two things are true at once, and keeping them separate is what makes the picture honest.
- Strong foundation. In AI fragrance answers, faces.ae is the #2 most-cited domain and the #1 retailer (ahead of noon, Namshi, Golden Scent and Sephora), at ~77% positive sentiment. Faces is strongest on Google's surfaces, AI Overviews 15.7% and Gemini 12.3%.
- Clear gap. Faces wins the transactional slots (where-to-buy, same-day delivery) but is near-absent from the recommendation slots, the "best oud", "reviews" and even "authentic oud" (2.7%) answers where AI tells shoppers what to buy.
- Read it right. The only names ranked above Faces are fragrance houses (Tom Ford, Maison Francis Kurkdjian), because they are the products being asked about. Among retailers, Faces is #2, neck-and-neck with Sephora.
So this is not a visibility rescue. It is a measurement-and-growth opportunity: prove the channel, then grow it.
~75%
targeted visibility uplift via Profound Agents
~40% / ~31%
UAE / KSA search demand already triggers an AI answer
The approach
We attack it the way the Lab attacks any data problem: instrument first, then prove.
- Establish the full baseline. Extend Profound's live fragrance prompts into Skincare and Arabic-native coverage (our prompt library is built for exactly this), so we measure Faces across the categories that matter, then size where the room to move is.
- Two measurement layers. A wide monitor set for coverage and a deep experiment set with a holdout control, so any uplift is provably ours, not background AI drift.
Partner: Profound
Profound is the category-leading platform for Answer Engine Optimization, the new SEO for AI search. It tracks how brands appear across ChatGPT, Gemini, Google AI Overview and Google AI Mode the way Semrush and GA track Google: measuring visibility, citations and sentiment, and running the agents that win the answers. It is trusted by retail and consumer brands including Sephora USA, Dyson and Lacoste. Setup is collaborative: we share the prompt library & baseline analysis; Profound's dedicated account team loads and tunes it.
POC: economics & success metrics
~$12K
3-month POC · paid upfront, net 30 · opt-out gate
2 + 2
countries (UAE, KSA) + languages (EN, AR)
4
AI engines: ChatGPT, Gemini, AI Overview, AI Mode
Free
agent credits during the POC · 10 competitors
| KPI | Baseline | Target |
|---|---|---|
| Visibility Score how often Faces appears in answers | Early read ~10.5% (fragrance, #2 retailer) | Full baseline, then ~75% uplift |
| Citation Share faces.ae share of AI citations | #2 domain · #1 retailer | Hold the lead; extend to skincare |
| Recommendation visibility "best", "reviews", "authentic" prompts | Near-absent (authentic oud 2.7%) | Win via agents on priority prompts |
| Sentiment / accuracy what AI says about Faces | ~77% positive | Maintain; correct any errors |
| AI-driven traffic visits from AI engines to faces.ae/.sa | Tracking being set up | Baseline AI-driven visits, then grow |
The 75% uplift
The Math: once the baseline is set across Perfumes + Skincare, Profound Agents target a ~75% visibility uplift (illustratively, 6.6% to 11.6%).
The Shift: we apply that lift to Faces' real baseline and the biggest-gap prompts, not a generic number.
Win the recommendation slots
The Math: Faces wins "where to buy" but is near-absent in "best oud", "reviews" and "authentic oud" (2.7%), the answers that tell shoppers what to buy.
The Shift: agents build the content and structure that get Faces cited in those answers.
Lean into our edge
The Math: Faces is strongest on Google's surfaces (AI Overviews 15.7%, Gemini 12.3%), and the #1 Saudi perfume question is authenticity, where Faces is an authorized retailer.
The Shift: prioritize the engines and topics where we already lead.
Economics, de-risked by design: the commitment is ~$12K for a 3-month POC (paid upfront, net 30), with agent usage free throughout. A 3-month opt-out (2 weeks' notice) is our clean go/no-go gate: we only continue the 12-month term (~$50K) if the POC proves the lift. No new headcount.
Roadmap & scaling
| Stage | Window | Scope |
|---|---|---|
| Establish baseline | Weeks 1-3 on infosec clearance | Stand up tracking across Perfumes + Skincare, both markets. Set the starting point and the gap map. |
| Optimize & prove | To the opt-out gate | Run agents on the biggest-gap prompts; measure visibility uplift vs holdout; stand up AI-driven traffic tracking. |
| Operationalize | Post-gate → Jun 2027 | Scale agents across the AI-visible surfaces (PLPs, blogs, Reddit, YouTube); expand categories. |
| Group playbook | 2027 onwards | Roll the validated GEO playbook across managed, JV (Shopify) and own-concept brands (Level, Tryano, Tanagra…). |
Operating model at scale. Ownership splits cleanly, governed by an audit layer:
- Harry's central digital-marketing team owns the analytics and builds the approved agent library.
- Brand teams (Faces, Level…) run those pre-designed agents human-in-the-loop to take action, only after the central audit layer signs off.
- Capability built in-house: we nominate 2 people for Profound University (a 2-week Marketing Engineer pod), so the team can build and run the agents themselves.
How it compounds: at scale, GEO analytics become an input to the Lab's product-enrichment pipeline (the Product Golden Record): the content signals that win AI citations feed how products get enriched, so AI-visible answers and storefront-ready PDPs reinforce each other.
Timing: once legal + infosec clears, baseline tracking and agent build run in parallel, anchored by an on-site agent-design workshop with Profound (targeting early July), then optimize and measure through to the opt-out gate.