Their team opens only the publisher domains that already passed.
Judging a publisher properly means opening the site, reading the traffic trend, checking authority against real traffic, scanning the outbound footprint and deciding whether the niche fits. At agency volume that work either does not happen or it happens badly. Horus now does the screening and writes each verdict into their own database.
- System
- Horus, publisher screening
- Scope
- Discovery and quality audit
- Live since
- February 2026
- Where verdicts land
- Their own Supabase
- Operator time per week
- Not measured yet
- Numbers on this page
- 1, with its source
Watch one record go through it.
Click a stage to see the real screen the team works in.
6 stages · 1 human gate · 7 vendors
drag · pinch · click a stage
- 01
Discovery
One brief, editable filters
Google
Ahrefs - 02
Vet and qualify
About 60 seconds per domain
DataForSEOAhrefs - 03
Find contacts
No purchased lists
In-house - 04Human gate
Held for approval
Your call on anything borderline
- 05
Send and classify replies
From your own inboxes
Instantly
OpenRouter - 06
Monitor links
Live, dofollow, indexed
Supabase
Marketing Signals run their own pitch and their own publisher relationships. This instance ends at the verdict.
The real screen · Discovery
The steps that went away.
By hand, per domain
- 01Open the site and read itperson
- 02Read the traffic trendperson
- 03Check DR against real trafficperson
- 04Scan the outbound link footprintperson
- 05Decide whether the niche fitsperson
- 06Log the verdict somewhereperson
Per-domain time is measured with the client at intake. Published only once measured.
With Horus
- 01Discovery fills the queue from a briefsystem
- 02The audit scores each domain and names the red flagssystem
- 03The verdict is written to their own databasesystem
- 04Their team opens what passedperson
Six manual steps became one review step. The count is the claim; the hours follow when measured.
What we can actually show.
One tile per figure, and every tile says where its figure came from. A pending tile stays pending until something is measured.
45,000+
domains screened in three months
Row count from a live account, May to July 2026. Rounded down.
Not measured yet
operator hours saved per week
Appears once an intake baseline and a 30-day reading exist. Never derived.
7 months
in production
Counted from the first production row, 3 February 2026. Updates itself.
Built, kept human, left alone.
Built
Discovery from a plain-language brief into editable filters
The quality audit: niche match, editorial signals, DR, traffic trend, market price
Verdicts and red flags written straight into their Supabase
Kept human, on purpose
The final call on any borderline domain
The pitch and the publisher relationship
Not built, and why
Outreach sending: they run their own process and it works
A new dashboard: the verdicts land where their team already looks
The gates the client signs.
- Week 1SetupKickoff, intake, access, architecture note
- Gate 1UI signoffMockup approved before a line is built
- Weeks 2 to 3BuildModules, internal QA
- Gate 2Client testingTheir team runs it on real domains
- Week 4Docs and walkthroughRecorded, handed over
- Gate 3AcceptanceThen it runs, monitored
Standard type-1 delivery. The teal diamonds are the points where the client signs, not us.
Months in production, and what shipped since.
7 months
no rebuild
A demo is easy. Month twelve is the product.
Since launch
- Jul 2026
Listicle outreach added: finds the best-of lists that rank competitors and pitches for inclusion
- Jul 2026
Link insertion scoring and link monitoring live: every indexed article rated, and placed links checked on live, dofollow and indexed
- Aug 2026
SEO data providers moved onto Tasken's keys; the customer holds only the AI key
Bring one workflow. Thirty minutes.
Tell us the job your team repeats and we will tell you whether a build is worth scoping. If it is not, we will say so.