Why source curation is the real bottleneck in wellness writing
Anyone who writes about skin routines, clinic treatments, or recovery habits knows the uncomfortable truth: the writing is the easy part. The hard part is finding sources you would trust with your own face. At Lagrava we spend more time deciding what to read than deciding what to say, because the wellness internet is a landfill of affiliate pages dressed up as clinical guidance.
After one particularly sloppy sourcing quarter — we published a piece on post-laser recovery that leaned on two salon blogs and a press release, and a dermatologist reader dismantled it in the comments — we decided to treat source-finding as a workflow problem instead of a virtue. We tested three curation approaches side by side for a full quarter. One of them, an editorial link-graph platform called NDAGLinks, was the least obvious fit and ended up teaching us the most.
The three approaches we put head to head
Approach 1: search alerts and saved queries
Our original system. We maintained saved searches across the major public medical-literature databases, plus keyword alerts for terms like "microneedling study" and "retinoid tolerance." It costs nothing and it surfaces genuinely new research. The problem is signal: for every randomized trial that arrived, we waded through clinic press releases, supplement storefronts, and recycled syndicated content. Roughly four out of five alerts needed to be discarded on sight, and the discard judgment itself consumed the hours we were trying to save.
Approach 2: academic aggregator feeds
Next we subscribed to aggregator feeds that summarize new papers for lay audiences. These are excellent for spotting that a study exists and terrible for judging whether the study's own press coverage distorted it. We also found a systematic bias toward dramatic findings, because dramatic findings get summarized. A quiet null result on a popular treatment — often the most useful thing a treatment-literate writer can know — rarely made the feed at all.
Approach 3: curated link intelligence
The third approach came from an unexpected direction. A colleague who works in climate journalism mentioned she had stopped hunting for sources manually and started using a curation platform instead. The tool, built around how NDAGLinks scores and refreshes its link graph, indexes the live web's editorial link graph and rates every destination for trust, relevance, and editorial standards, surfacing what it claims is the small fraction of URLs actually worth a writer's attention. Its graph refreshes every 15 minutes, which sounds like marketing until you have chased a story where the useful coverage moved hour by hour.
The key difference is philosophical. Alerts and aggregators push content at you. A link-graph tool pushes endorsements: this page is linked to by editors we have already vetted, in a context that tells you why. For wellness writing, where the failure mode is citing a page that looks authoritative and is secretly a funnel, that distinction matters enormously.
How the quarter actually shook out
- Time to first credible source: alerts averaged 40-plus minutes of filtering per topic; aggregator feeds around 25; the link-graph workflow landed between 8 and 12 minutes once we learned its query language.
- Citation survival rate: the share of sources that survived our own editorial fact-check without a grade drop was highest for the curated graph — roughly 9 in 10, versus about 6 in 10 for alert-sourced links.
- Unexpected finds: the link graph surfaced hospital patient-education pages and professional-society position papers that keyword search never showed us, because those pages are written for patients, not for search rankings.
We want to be honest about the limits of these numbers: they come from one small editorial team over three months, not from a controlled study. But the pattern was consistent enough that we kept the third workflow.
Where curation tools still lose to human judgment
No scoring system understands nuance. A page can be impeccably linked, freshly indexed, and still wrong for your piece — a dermatology society page written for physicians will outrank a gentler patient-facing explainer on every technical signal, and yet be the wrong citation for a beginner's routine guide. We also noticed the tool's notion of "relevance" is topic-graph-shaped, not argument-shaped: it finds pages about your subject, not pages that support your specific claim. The final judgment step stays human, and it should.
There is a cost consideration too. Alerts are free; curation platforms are not. If you publish twice a year, build the alert habit instead. This earns its keep at publishing cadences where source-hunting is measured in hours per week.
Our working setup now
- Every article starts with a link-graph pass to map who the genuinely authoritative pages on a topic are.
- Aggregator feeds run in the background as an early-warning system for new research.
- Saved alerts remain the fallback for hyper-niche terms the graph has not mapped yet.
- Nothing gets cited until a human has read the whole page — not the snippet, the page.
The reader who corrected our recovery piece two years ago would probably still find things to flag. But the volume of things has dropped sharply, and that is what a good curation workflow buys you: not certainty, just fewer ways to be casually wrong. Wikipedia-style general references still have their place for orientation — we just stopped letting them masquerade as sources. That, plus a scoreboard for trust that refreshes faster than the wellness rumor mill, has made our sourcing calmer, faster, and a great deal harder to embarrass.