I posted an outline of vector search the other week and shortly afterwards became aware of some work Google has been involved in: Accelerate medical research with PubMed data now available in BigQuery. In essence, Google has created a vector-based search system over PubMed Central.

We set up a small test-bed to explore whether this approach could be incorporated into Beyond Trip. We’ve been testing it using clinical questions, traditional keyword searches, and queries that sit somewhere in between – and the results have been genuinely impressive.

One search in particular, “Creatine use and cognitive ability”, clearly demonstrated the value of vector search. When we ran this query through the test-bed, the top result was The effects of creatine supplementation on cognitive function in adults: a systematic review and meta-analysis.

I then repeated the same search in Trip using the identical wording (Creatine use and cognitive ability), and that paper was not retrieved. However, when I reran the Trip search using “cognitive function” instead of “cognitive ability”, the paper appeared immediately. The content of the paper is the same in both cases – only the phrasing of the query changed. So, seamlessly the vector search was able to ‘understand’ that cognitive function and cognitive ability are virtual synonyms.

What this neatly illustrates is how vector search moves beyond literal term matching and begins to reflect clinical meaning. By recognising that “cognitive ability” and “cognitive function” are effectively synonymous, vector search bridges the gap between how clinicians think and how evidence is described in the literature. For tools like Beyond Trip, this has the potential to reduce missed evidence, lower the cognitive burden of searching, and make high-quality research easier to find – even when the wording doesn’t line up perfectly.