# True South — what agents see

> See what AI agents say about your business, which sources they use and what you can change. Australia and New Zealand.

Canonical: https://truesouth.cc/
Updated: 2026-11-22

## What the audit does

Agents are now part of the first pass: they shape who gets considered. They work from what they can find, read and trust. The audit runs real agent tasks against the business and records what comes back: the answer, the source, and whether the cause is your own page, a third party, or the model’s prior.

Agent readiness is the underlying discipline: making the business legible, trustworthy and actionable to the agents your buyers already use.

It is not GEO or answer-engine optimisation. That sells visibility in AI answers. Agent readiness sells capability: whether an agent can find you, read the facts it needs, verify your claims, and complete an action without a human filling in a form.

## The test

Ask an assistant to shortlist three suppliers in your category, then check:

1. **Find.** Can it find you? Crawl rules, a sitemap, structured data and links an agent can follow without a browser.
2. **Read.** Can it read you? Markdown, semantic HTML, and the facts that matter — services, prices, proof — outside the JavaScript.
3. **Trust.** Can it verify you? A named entity, claims with evidence, a security contact, and pages that say when they were last checked.
4. **Act.** Can it act? A contact path, a booking, an API or a way to buy without a human in the loop.

If the answer is no, you are invisible to the fastest-moving part of the funnel.

## Why now

An agent is software that does a job for someone, around the clock. The one many buyers already use is the AI chatbot they ask to find and compare suppliers.

- 51% of B2B software buyers start their research with an AI chatbot more often than with Google, up from 29% in April 2025.
- 69% chose a different software vendor than they had planned, on the chatbot's guidance.
- Fewer than 15 of the 200,000 most-visited sites give an agent a way to act.

Sources: G2, March 2026 survey of 1,076 B2B software buyers (<https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html>); Cloudflare, April 2026 (<https://blog.cloudflare.com/agent-readiness>).

> "We are having to shift incredibly quickly from a marketing perspective, as we have to market to humans and to agents." — Mim Haysom, executive general manager of brand and customer experience, Suncorp, quoted in the Australian Financial Review, 21 Nov 2026 (<https://www.afr.com/technology/ai-shopping-agents-have-arrived-and-major-brands-are-quietly-panicking-20261121-p61fo9>).

Agents already shape who gets considered. Most businesses have yet to see how they look to one, and few know what to change. We show you both. Now is the time to look, while it is still early.

## Offers

All prices in AUD, excluding GST, valid to 30 June 2027.

### Agent Readiness Scan — 4,900, fixed price

Ten business days. 25 checks across five dimensions:

- **Discover:** robots.txt AI rules, Content Signals, sitemap, Link headers, llms.txt.
- **Read:** markdown negotiation, no-JS content parity, semantic structure, JSON-LD, token weight of the five most important pages.
- **Verify:** legal entity and ABN (where one exists), contact facts, security.txt, evidence for claims, freshness and versioning.
- **Act:** agent-usable contact path, booking, API catalog, MCP server card, OAuth discovery.
- **Control:** AI bot rules, Web Bot Auth readiness, rate limiting, access logging, commerce readiness.

Plus: five buyer prompts across three assistants; a benchmark against two named competitors; a plain-English judge's note; an effort × impact gap map; a 45-minute recorded readout; 14 days of questions; and a day-30 re-scan.

Deliverable: a 12–16 page report with findings, evidence and a prioritised gap map, plus `scan.json` and the raw agent transcripts. The client owns all of it.

Executable remediation prompts and implementation belong to the Remediation Sprint. The scan also excludes content writing, paid media, model selection, security testing, legal advice and ongoing monitoring.

Founding cohort: the same scope at 2,450 for the first five scans in exchange for a named quote or an anonymised case note.

### Remediation Sprint — fixed scope

- **S:** 18,000, three weeks. One surface, one decision.
- **M:** 26,000, four weeks. Up to three surfaces or business units.
- **L:** 34,000, five to six weeks. Whole-of-organisation, board-ready.

We ship the priority fixes: content machines can read, facts they can verify, and paths they can act on. Deliverable: working fixes, shipped and re-tested.

### Agent interface pilot — by quotation

Two weeks, one action an agent can take — a booking, an enquiry, a price check — on the systems the client already runs. Scoped after the scan. A larger programme only if the pilot earns it. Deliverable: one action an agent can take, working in two weeks.

### Quarterly re-scan — by quotation

The same checks four times a year, with a delta and one advisory session. Scoped after the scan. No monthly tiers.

## Method

1. **Scan** — measure how agents find, read, verify and act on the business today. Ten business days.
2. **Fix** — ship the priority fixes and re-run the scan to prove they worked. A Remediation Sprint, three to six weeks.
3. **Pilot** — one action an agent can take on the systems you already run. Two weeks, quoted after the scan.
4. **Re-scan** — the same checks each quarter, with what changed. Quoted after the scan.

The scan and the sprint are fixed price. The pilot and the re-scan are quoted after the scan.

## The gap

Cloudflare scanned the 200,000 most visited domains in April 2026:

- 4% publish Content Signals.
- 3.9% serve markdown when an agent asks for it.
- Fewer than 15 sites in the dataset publish an MCP server card or an API catalog.

Most robots.txt files are written for search crawlers, not agents. The gap is the opportunity.

Source: <https://blog.cloudflare.com/agent-readiness>

## Measured

Two scores, two methods. Both are labelled in <https://truesouth.cc/scores.json>; do not compare them without naming the method.

**True South 25-check method (self-assessment), 19 Nov 2026: 66/100, Legible.**

- Discover 100
- Read 100
- Verify 40
- Act 62
- Control 40

**Cloudflare public 23-check scan, 19 Nov 2026: 47/100, Level 4 (Agent-Integrated), up from 27.**

- Discoverability 3/4
- Content 1/1
- Bot access 2/2
- Protocols 1/8
- Commerce optional

Source: <https://isitagentready.com>.

## MCP

A read-only MCP server is live at <https://truesouth.cc/mcp> (Streamable HTTP, JSON responses). Server card: <https://truesouth.cc/.well-known/mcp/server-card.json>.

Tools: `get_services`, `get_pricing`, `get_method`, `get_score`, `get_contact`. No authentication, no writes.

## The sky

The chart at <https://truesouth.cc/sky> is real astronomy, plotted for the observer’s latitude and time. The Cross, the Pointers, the pole — and how true south is found by construction.

## About

Tamir Levin is the founder of True South, in Melbourne. Tamir has led technology for more than thirty years as a CIO, CTO and a strategy adviser. He still builds things himself, including the tests behind every report, and he is the person you talk to.

[Portrait of Tamir Levin](https://truesouth.cc/tamir-levin.jpg). This site is the first case study: its own scores, changes and failures are published as they are measured.

- Market: Australia and New Zealand, working remotely across the southern hemisphere.
- No ABN or legal entity details are published yet.
- No client logos, testimonials or invented metrics appear on the site.

## Contact

`hello@truesouth.cc`. The Book button on the site opens an email to this address.

To start: name the domain, the category you compete in, and two competitors you would hate to lose to.

## Machine-readable

- <https://truesouth.cc/pricing.json>
- <https://truesouth.cc/scores.json>
- <https://truesouth.cc/sample-report>
- <https://truesouth.cc/sample-report.pdf>
- <https://truesouth.cc/.well-known/mcp/server-card.json>
- <https://truesouth.cc/llms.txt>
- <https://truesouth.cc/sitemap.xml>
- <https://truesouth.cc/.well-known/security.txt>
