True South (this site)
https://truesouth.tamirlevin.dev/
True South (this site) scores 66/100, which puts it at Legible. Strongest: discover (100). Weakest: verify (40).
The Cross, with its axis extended to the south celestial pole. Real positions, checked by construction.
True south is the real direction,
not the convenient one.
What we found
Across 25 checks, 14 passed, 5 were partial, 5 failed and 1 did not apply. The three fixes below carry the most weight; each is scored against a written pass condition, and the evidence for every result is in the dimension sections.
This is True South's own site, measured with the paid 25-check method. The public Cloudflare scan (47/100) is published at /scores.json; the two numbers use different methods and are not compared.
Scroll across the table to see all counts.
| Dimension | Question | Score | Pass | Partial | Fail | N/A |
|---|---|---|---|---|---|---|
| Discover | Can an agent find you? | 100 | 5 | 0 | 0 | 0 |
| Read | Can it read you? | 100 | 5 | 0 | 0 | 0 |
| Verify | Can it verify you? | 40 | 1 | 2 | 2 | 0 |
| Act | Can it act? | 62 | 2 | 1 | 1 | 1 |
| Control | Can you govern it? | 40 | 1 | 2 | 2 | 0 |
14 pass · 5 partial · 5 fail · 1 n/a
Fix these first
Three changes, ordered by severity. Each is scored against a written pass condition; the evidence sits in the dimension sections.
V1 — Named entity or principal fail · critical
Pass means: A named entity or person, plus a location, is on the site.
Evidence: Brand only; no legal entity, ABN or named principal · Founder-led is stated but unnamed
A1 — Agent-usable contact path partial · critical
Pass means: A form endpoint, email or booking URL that works and is monitored.
Evidence: Form writes a local draft; no endpoint receives it · Email domain not registered
V2 — Working contact facts partial · critical
Pass means: At least two independent, working channels, consistent across the site.
Evidence: hello@truesouth.cc not live; form copies a draft; no second channel · Blocks agent and human contact
A partial counts as half a pass. Fixing all three moves the score more than any other combination of checks.
Discover — 100/100
Can an agent find you?
All 5 checks passed. This is the strongest part of the surface.
What good looks like: Agents find the site through robots, a sitemap, structured data and a machine index.
| ID | Check and evidence | Status |
|---|---|---|
| D1 | robots.txt is present and valid critical · auto /robots.txt 200, valid syntax, wildcard group and sitemap reference No named AI crawler rules; that is tested separately as C1 | pass |
| D2 | Sitemap exists and is referenced high · auto /sitemap.xml 200, one URL, referenced from robots.txt | pass |
| D3 | Discovery Link headers medium · auto Link: alternate text/markdown, describedby llms.txt, service-desc pricing.json | pass |
| D4 | Entity and service structured data high · auto JSON-LD ProfessionalService plus four Service offers; parsed cleanly | pass |
| D5 | Machine index low · auto /llms.txt and /index.md 200, links resolve | pass |
Read — 100/100
Can it read you?
All 5 checks passed. This is the strongest part of the surface.
What good looks like: The same facts are available as HTML and as markdown, without JavaScript.
| ID | Check and evidence | Status |
|---|---|---|
| R1 | Markdown negotiation high · auto Accept: text/markdown returns text/markdown with Vary: Accept | pass |
| R2 | Key facts without JavaScript critical · manual Services, prices and contact are in the HTML; chart degrades without JS | pass |
| R3 | Semantic structure high · auto One h1, ordered headings, landmarks; Lighthouse accessibility 100 | pass |
| R4 | Token weight of key pages medium · auto index.md 5.5 KB, about 1.4k tokens | pass |
| R5 | Freshness and versioning low · manual Score dated 19 Nov 2026; prices valid to 30 June 2027 | pass |
Verify — 40/100
Can it verify you?
1 of 5 checks passed. The largest gap is named entity or principal (fail), worth about 20 points here.
What good looks like: A named entity, working contact channels, evidence for claims and external corroboration.
| ID | Check and evidence | Status |
|---|---|---|
| V1 | Named entity or principal critical · manual Brand only; no legal entity, ABN or named principal Founder-led is stated but unnamed | fail |
| V2 | Working contact facts critical · manual hello@truesouth.cc not live; form copies a draft; no second channel Blocks agent and human contact | partial |
| V3 | Claims have evidence high · manual No case studies or testimonials; no fabricated claims either Evidence is the site's own measurements | partial |
| V4 | security.txt medium · auto /.well-known/security.txt 200, expires 2027-11-19, URL contact | pass |
| V5 | External corroboration medium · manual No registry, industry body or LinkedIn profile linked | fail |
Act — 62/100
Can it act?
2 of 4 checks passed. The largest gap is booking or response commitment (fail), worth about 25 points here.
What good looks like: A price, a contact path, and a booking or API an agent can complete.
| ID | Check and evidence | Status |
|---|---|---|
| A1 | Agent-usable contact path critical · manual Form writes a local draft; no endpoint receives it Email domain not registered | partial |
| A2 | Prices or ranges published high · auto Scan AUD 4,900 + GST; sprint tiers; retainer tiers; validity stated | pass |
| A3 | Booking or response commitment high · manual No booking link and no stated response time | fail |
| A4 | API, catalog or MCP endpoint medium · auto /mcp 200, five read-only tools; server card live | pass |
| A5 | Agent commerce readiness low · manual No online goods; advisory engagements only | na |
Control — 40/100
Can you govern it?
1 of 5 checks passed. The largest gap is rate limits and abuse controls (fail), worth about 20 points here.
What good looks like: Explicit crawler rules, content signals, and limits on anything public that can be abused.
| ID | Check and evidence | Status |
|---|---|---|
| C1 | Explicit AI bot rules high · auto Wildcard allow plus Content-Signal; no named AI crawler rules | partial |
| C2 | Content Signals medium · auto Content-Signal: search=yes, ai-input=yes, ai-train=yes | pass |
| C3 | Rate limits and abuse controls medium · manual No rate limit or abuse control on /mcp or future /api/check | fail |
| C4 | Agent traffic is observable low · manual Cloudflare logs available; no documented agent-traffic method or privacy note | partial |
| C5 | Agent identity policy low · manual No stated policy on signed agents or Web Bot Auth | fail |
What agents say
The full scan records five buyer prompts across three assistants, with answers recorded verbatim.
Not included in this run.
Competitor benchmark
The full scan runs the same prompts and checks against two named competitors.
Not included in this run.
Method and limits
This report measures the surface an AI agent meets when it arrives: whether it can find the business, read the facts, verify the claims and complete an action. Each check has a written pass condition, and every result carries its evidence.
- 25 checks across five dimensions: discover, read, verify, act and control.
- Weights: discover 15%, read 20%, verify 25%, act 25%, control 15%.
- Status values: pass, partial (half credit), fail, or n/a (excluded and renormalised).
- Stages: Blind (<25), Visible (<50), Legible (<75), Verifiable (<90), Actionable (90+).
Limits: point-in-time; no penetration testing; no legal advice; no model fine-tuning; results reflect what a public request can see, not what an authenticated user or a signed agent sees.
Next
Fix the three priorities, then re-scan at day 30. The re-scan is included in the engagement. If the fixes need shipping rather than describing, the Remediation Sprint takes the top of this list and does the work.
truesouth.tamirlevin.dev · contact form