The starting point
A lifestyle publisher with 3M monthly pageviews and a stagnant Discover surface came to us in February 2026. They were getting 3M Discover impressions a month — decent, but flat for two quarters. Their team felt like they'd hit a ceiling. We thought the ceiling was self-imposed.
The audit
We pulled 90 days of Discover data from Search Console and segmented by author, topic, image style and headline pattern. Two patterns jumped out: articles with original photography earned 3x the impressions of articles with stock photos. Articles with specific, numerical headlines (e.g. '7 ways to fix X') earned 2x the impressions of generic headlines (e.g. 'How to fix X').
The image investment
We commissioned an in-house photo team — one full-time photographer, one part-time stylist. Every article published from March onwards used original photography. The upfront cost was significant, but the math worked: the photo team paid for itself within 6 weeks at the publisher's existing CPM rates.
The headline framework
We rewrote every headline using a strict framework: specific number + concrete benefit + curiosity hook. The headlines were 20% longer than before but performed dramatically better in Discover. We A/B tested the new headlines against the originals on social and saw a 3x click rate uplift, which predicted (correctly) the Discover lift.
The publishing cadence
We doubled the publishing cadence from 4 articles a day to 8, but only on the topics where Discover was already sending impressions. Topical concentration matters in Discover — the algorithm builds confidence in you for specific topics, not broadly. By doubling down on the topics that worked, we accelerated the confidence-building loop.
The technical fixes
Three technical fixes contributed: enabling max-image-preview:large in robots meta (forgotten on the publisher's CMS template), shipping LCP under 2s on mobile (was 3.4s), and adding Article schema with author and publisher fields (was missing on 40% of articles). None glamorous, all measurable.
Month-by-month results
March: 5M impressions (+67%). April: 8M (+167%). May: 12M (+300%). Three things compounded — better images attracted clicks, higher CTR signalled quality to Google's algorithm, the algorithm increased the surface area allocated to the publisher. The growth was non-linear because the inputs reinforced each other.
What we'd warn against
Discover traffic is volatile. A single algorithm tweak can cut your impressions by half overnight. Don't build your business on Discover alone — use it as an amplifier on top of a base of organic search traffic and direct loyalty. The publishers that survived the 2023 and 2024 Discover algorithm shifts were the ones with diversified traffic, not the ones who optimised exclusively for Discover.