Our impact

Google Ad Grants is the single largest reason anyone finds this site. It places our pages in front of people searching for exactly the thing we have written about - a BI-RADS category, a Gleason score, what “correlate clinically” means on a report.

It is worth being precise about what that grant is. It is a ceiling on donated advertising, not money we spend and not money we receive. It also comes with conditions: a minimum click-through rate, conversion tracking, and policy compliance, with suspension if those slip. Health advertising carries extra restrictions on top.

We are not close to using the full ceiling. That is a statement about how much qualifying demand we have found so far, not a claim about how much we are doing.

The rule we hold ourselves to is that a medical claim has to trace to a primary source - the National Cancer Institute, CDC, FDA - and that we keep the source's hedging rather than tidying it into something more confident than the evidence.

The report-decoder pages are the part that gets used most. They exist because of a pattern in what people search for: not “prostate cancer” but “gleason score 8”. Somebody is holding a piece of paper with a number on it, and they want to know what the number means before their next appointment.

We do not tell anyone what will happen to them. A page explains what a value describes and what to ask about it. Predicting an individual outcome from a number on a report is something no website can honestly do.

Each claim-check takes a specific claim - fenbendazole, alkaline diets, frankincense oil, “sugar feeds cancer” - and works out what the actual evidence says, with sources. We say plainly when something has no good evidence behind it, and equally plainly when a claim has a grain of truth that is being oversold.

We also check ourselves. Automated checks run over every page before it publishes, looking for claims that have drifted from their source, guidance that contradicts itself, and escalation advice that could be misread. Those checks have caught real errors in our own work, and when they do we fix them and say so on our updates page.

Coverage is uneven and we would rather say so than round it up. Five languages have substantial libraries - Vietnamese, Chinese, Arabic, Russian and Thai each have between 800 and 2,000 pages. The other twelve have a foundational core of around 30 pages each.

Machine translation got us that coverage. It also, at least once, turned a five-year survival figure of 67.8 percent into 567.8 percent on a page about somebody's cancer coming back. We found it, corrected it against the original source, and built an automatic check that now compares every number in all 7,746 translated pages against its English original before anything publishes.

That is the honest state of it: broad reach, machine-assisted, with human review as the gap we are still closing. Bilingual reviewers are one of the things we most need.

We have written thousands of sourced pages that almost nobody can find through a search engine. Paid placement through the Ad Grant delivers many times more readers than organic search does, which is the wrong way round for a library that is meant to last.

We have ruled out the easy explanation. The unindexed pages are not thin or low quality; we sampled them and they are substantial, well-sourced articles that have simply never been crawled. The likeliest cause is that a new domain has not yet earned enough authority for Google to spend crawl budget on it.

What that means practically: the work that would help most right now is not more articles. It is earning the kind of citations and links that make a search engine treat a site as worth reading. That is slow, and it is the thing we are worst at.