Skip to content
Now

What has our attention at the moment.

A dated snapshot rather than a positioning statement. Reviewed every six weeks; if the date below is older than that, treat what follows with suspicion.

Now

Current focus

Four things, as of .

Written for someone deciding whether our current preoccupations overlap with their problem. If none of these sound like your situation, that is useful information too.

01

Measuring AI discovery without pretending it is solved

Most of what is sold as GEO measurement counts mentions and calls it visibility. We are building protocols that separate three different questions — whether a source gets retrieved at all, how it gets used once retrieved, and whether any of it produces a visit — because the published research only supports confident claims about the middle one.

02

Offline revenue signal as the default, not an upgrade

Automated bidding optimises toward whatever signal it is given, which means a campaign fed form-fill counts will produce form fills faster than a human could. Getting closed revenue back into the platforms is now the first build on most paid engagements rather than a phase-two item.

03

Freight dispatch as an operated service line

Eight dispatch and back-office services are live on the site. The line is operations work rather than marketing, and it is being run as such — the growth service attached to it is the smaller half.

04

Market coverage past the English-first tier

120 markets have their own pages, and the next tranche is the one that needs the most care: markets where English-language content will not carry domestic visibility and saying otherwise would be selling a translation as a strategy.

Standing work

The parts that do not change.

The editorial library runs to 232 long-form guides across 30 service areas, each with visible sources, a review date and stated limitations. Claims with short shelf lives — platform behaviour, legal requirements, current research — get checked more often than durable concepts.

Citations are re-verified periodically rather than on publication alone. The last full pass found one genuine problem worth fixing: a source host that reset the TLS handshake, so five citations failed silently for every reader who clicked them.