The Test Premise
Does adding a creator face to a food-only cooking thumbnail actually increase CTR? The conventional wisdom says yes — faces drive clicks. But food photography is also compelling in its own right, and the cooking niche has a long tradition of food-only thumbnails that perform well.
This case study documents an 8-video systematic A/B test run by a cooking channel (44,000 subscribers) to measure the face vs. food-only CTR difference in their specific niche and audience.
Channel Context
The channel had historically used food-only thumbnails — high-quality close-up dish photography with a hook text overlay. Average CTR across the library: 4.2%. This is above the platform average but below the 6–8% range achievable by the highest-performing cooking channels.
The channel owner observed that the highest-CTR videos in the cooking niche on YouTube consistently showed a creator face alongside the food, while food-only thumbnails were more common among lower-CTR channels. The question: was this correlation causal, or were other factors (bigger channels, better search optimisation, more viral topics) explaining the difference?
Test Design
For 8 consecutive videos, the channel created two thumbnail versions per video:
Version A (food-only): Close-up of the finished dish, hook text overlay, consistent warm colour treatment. The channel's existing style.
Version B (face + food): Creator face occupying 35–40% of the left side of the frame, delighted or surprised expression directed toward the dish. The dish occupying the right 50% of the frame. Same hook text.
Each video published with Version A. After 7 days (once initial traffic normalised), switched to Version B for 7 days. Recorded CTR for each 7-day window. All other factors (title, tags, publish time, video length) held constant.
Results Across 8 Videos
| Video | CTR Version A (food only) | CTR Version B (face + food) | Lift |
|---|---|---|---|
| Pasta carbonara | 4.8% | 7.3% | +52% |
| Sourdough bread | 5.1% | 8.2% | +61% |
| One-pan chicken | 3.9% | 6.4% | +64% |
| Japanese curry | 4.2% | 7.1% | +69% |
| Banana bread | 6.3% | 8.8% | +40% |
| Homemade ramen | 3.7% | 6.9% | +86% |
| Beef stew | 4.1% | 5.8% | +41% |
| Chocolate cake | 5.6% | 9.1% | +63% |
| Average | 4.7% | 7.5% | +60% |
Confounding Factor: Time Effect
A potential confound in the manual swap design: video CTR typically declines as a video ages (older videos compete with fresher recommendations). Version B was always the second period — meaning it had the disadvantage of running on an older video.
If CTR naturally declined over the test window, the true face + food lift might be even larger than the 60% measured here, since Version B overcame the natural decay effect. This is a limitation of the non-randomised test design that future tests could address with YouTube Studio's native A/B test feature, which splits impressions concurrently.
Expression Analysis
Not all face thumbnails performed equally. Post-test analysis of the individual video results showed a pattern: the videos with the largest lift (ramen: +86%, Japanese curry: +69%, one-pan chicken: +64%) used more extreme creator expressions — jaw-drop for the ramen reveal, wide-eyed delight for the curry, hands-to-face surprise for the chicken.
The two videos with the smallest lift (banana bread: +40%, beef stew: +41%) used more moderate creator expressions — a pleased smile for the banana bread and a satisfied nod for the stew. Both were positive expressions, but neither was extreme by thumbnail standards.
This suggests that in cooking thumbnails, the lift from adding a face is modulated by expression intensity. A face with a genuinely extreme, positive food reaction outperforms a face with a moderate reaction — and the difference between extreme and moderate expressions may account for much of the variation across cooking channel CTR data.
Conclusions
Face + food thumbnails significantly outperform food-only thumbnails in this channel's audience. A 60% average CTR increase — moving from 4.7% to 7.5% — represents a major improvement that the channel maintained after the study concluded (switching the full library to face + food format).
Expression intensity modulates the effect. For the maximum CTR benefit, the creator's reaction to the food should be extreme — not politely positive. The emotional signal that food is genuinely exceptional (not merely good) is a key part of why face thumbnails outperform food-only versions.
Even the smallest lift (40%) was substantial. In a context where 4.7% was the baseline, a 40% lift (to 6.6%) still represents a major improvement. Even with moderate expressions, adding a face to food thumbnails appears to be a net positive for this audience segment.