I Scanned My Face with Pretty — Here's What Changed

Why I tried Pretty

I'd been wearing the same foundation shade for three years and the same lipstick for two — not because they were perfect, but because choosing alternatives felt overwhelming. AI beauty tools promised to remove that decision friction. Pretty (Fashionaholic's beauty analyzer) appeared in my feed multiple times. The pitch was simple: upload a photo, get personalized beauty recommendations in 30 seconds. I decided to test it for 60 days, then write about what actually happened. The broader question I was testing: do AI beauty tools deliver real, actionable insights, or are they just sophisticated marketing for product recommendations?

The initial analysis

The scan took 90 seconds — uploading a clean-face photo, answering 3 to 4 demographic questions, waiting for the analysis. The results came back as a profile across several dimensions.

• 01

  Skin undertone: cool (with neutral leaning). This contradicted my long-time assumption that I was warm-toned.

• 02

  Color season: Deep Winter. Aligned with the cool undertone finding.

• 03

  Eye shape: hooded with slight downturn.

• 04

  Face shape: heart with strong jawline.

• 05

  Skin condition: combination, slight dehydration, mild texture in T-zone.

• 06

  Recommendations: specific foundation shades from 8 brands, lipstick shade families to favor, eye makeup techniques for hooded eyes, hairstyles that balance heart face shape.

What I changed first

The biggest immediate change was foundation. If the cool-undertone analysis was right, my warm-toned foundation had been wrong for three years. I bought small samples of three cool-toned alternatives Pretty had recommended.

• 01

  Foundation: switched from warm-toned MAC Studio Fix to cool-toned Charlotte Tilbury Magic Foundation in a recommended shade. Immediate visible improvement — less orange, more 'looks like my skin.'

• 02

  Lipstick: shifted from warm coral tones to recommended cool berry tones. The difference was subtle but consistent — colors looked more 'right' rather than slightly off.

• 03

  Eye makeup: applied recommended hooded-eye techniques (avoiding heavy shadow above the crease where it disappears, using darker liner along upper lash line). Improvement in how my eye makeup actually showed.

• 04

  Total cost of changes: $180 for new foundation, lipstick samples, and one eye palette.

What surprised me

Two things were genuinely surprising about the experience.

• 01

  The undertone correction. I'd been wearing warm-toned foundation for years, presumably because warm tones suited my hair color (chestnut). The AI correctly identified that my actual skin undertone was cool — and the foundation change made a visible difference. I'd been making decisions based on wrong information.

• 02

  The hooded-eye recognition. I'd known I had hooded eyes but hadn't applied makeup techniques specific to the shape — using techniques meant for differently-shaped eyes, which is why my eye makeup never quite worked. The shape-specific recommendations were practical and immediately useful.

What didn't work

The hairstyle recommendations were the weakest part of the analysis. The suggestions felt generic — common cuts that 'balance a heart-shaped face' without specific brand or stylist guidance. I didn't act on these. The skincare recommendations were practical but obvious. Hydration focus, gentle exfoliation, SPF. Useful confirmation but not new information for anyone with basic skincare knowledge.

After 60 days

Sixty days in, the meaningful changes that stuck were: cool-toned foundation (definitely better than my previous warm-toned), cool berry lipsticks (better than warm coral), hooded-eye makeup techniques (better eye makeup application). Total: 3 meaningful behavior changes from the AI analysis. The estimated value: roughly equivalent to a $200 to $300 professional makeup consultation, delivered in 90 seconds for free. The analysis isn't as deep as a 2-hour session with a master makeup artist, but the cost difference (free vs $200+) makes the tool genuinely useful for the majority of users who'd never see a professional consultation otherwise.

Verdict and recommendations

Pretty (and AI beauty analyzers more broadly) deliver real, actionable value for most users. Three takeaways from the experiment:

• 01

  AI beauty tools work best for fundamentals (undertone, color season, basic face/eye shape analysis) — clear improvements over the 'guess and check' approach most users default to.

• 02

  Don't expect AI to replace human expertise for complex situations — special occasions, dramatic transformations, or unusual features still benefit from professional consultation.

• 03

  The 90-second tool delivers roughly 60 to 80% of the value of a $200 to $300 professional consultation. For most users, that's enough to justify regular use.