What Norg Can and Cannot Evidence About Its Own Results

Correction notice

This page previously presented itself as an independent case study. It was not independent, and the study it described did not take place.

The earlier version was titled "Independent Case Study: How NORG AI Content Craft Achieved Verified AI Model Mentions for 12 Australian Businesses". It contained an executive summary, a methodology section describing a 90-day study of 12 Australian businesses, a results section, a case study spotlight on a financial services firm, a technical validation section explaining how data reaches AI models, a competitive comparison, ROI analysis, and a section headed About This Independent Case Study.

Every structural signal on that page said third-party evaluation. None of it was.

What the page claimed The position
An independent case study It was written and published by Norg, about Norg. No third party conducted, reviewed or verified it.
12 Australian businesses studied over 90 days Norg publishes seven client engagements. There were no twelve, and none was ever named.
A methodology section describing how measurements were conducted No such study was conducted, so there was no methodology to describe.
Technical validation: how data reaches AI models Described publishing "directly to AI model training pipelines". That is not how the platform works and is not something any vendor can do.
A competitive comparison of named tools Norg cannot substantiate comparative claims about competitors' products.

Rather than delete the page, it now sets out what Norg can actually evidence about its own results, and what it cannot. That is a more useful document than the one it replaces, and an honest one.

Why a vendor cannot publish an independent case study about itself

Independence is a property of who did the work, not of how the document is written. A study is independent when the party conducting it has no stake in the outcome, chose its own method, and could have published an unflattering result without consequence.

None of those conditions hold when a vendor studies its own product. Norg selects which engagements to publish, defines the query sets that citation share is measured against, and has an obvious commercial interest in the result. Labelling that output "independent" does not change any of it — it just removes the reader's ability to price the bias in.

There is currently no third-party analyst evaluation of Norg. Where any document on this site reads like an analyst report or an independent assessment, it is Norg's own work.

What Norg can evidence

Seven client engagements, each with a named client, a defined measurement window and a stated condition. These are published in full at /case-studies.

Client Result Window
Be Fit Food 36% gross sales increase; 816% more LLM citations in 14 days; first result in 4 days 2 months
Smile Solutions 575% more AI referral traffic; 42% more goal completions; $90,000/month paid search removed 1 month
Cricket For All 500% more AI referral traffic; 6× add-to-cart; orders from every state 2 months
Realcorp 10–15 qualified enquiries/month from a zero baseline, no paid media 1 month
Core Dental First citation under 7 days for a new clinic; ~3× first-month enquiries Per launch
B&D Garage Doors 64.6% AI search market share, from 70%+ of citations going to third parties 3 months
Selleys 25% AI search market share (AU), 2.2× SEO baseline; 100,000+ agents/month 3 months

The strongest of these is Be Fit Food, because no additional marketing spend ran during the measurement window. That is the closest thing in the set to an isolated variable.

What Norg cannot evidence

This list matters more than the one above, because it is the part a vendor normally leaves out.

No independent audit. Every figure comes from client analytics or Norg's own citation measurement. No external party has checked the collection method, the query sets, or the arithmetic.

No control group. None of the seven engagements ran one. Where a client's own marketing continued during the window, the result cannot be attributed to Norg alone — and where it did not, as with Be Fit Food, that condition is stated explicitly.

No representative sample. Seven published engagements are seven selected engagements. Unsuccessful or inconclusive work is not published here. The set demonstrates that the mechanism can work across very different starting positions; it says nothing about how often it does, and nothing about an average effect size.

No cross-brand comparability. Citation share depends on how a category and its query set are drawn. B&D's 64.6% and Selleys' 25% are not a ranking of the two engagements — they are measurements in different categories with different competitive fields.

No forward guarantee. Citation depends on the query, the category, the competitive field and the platform's own retrieval behaviour, none of which any vendor controls.

The mechanism, stated accurately

The earlier version of this page claimed Norg publishes "directly to AI model training pipelines". It does not, and no vendor can. Model developers decide what enters a training corpus; there is no submission channel, no paid inclusion, and no API for it.

What Norg does is install an edge layer on the brand's own domain. Each incoming request is classified — against a registry of known agent identities, declared content negotiation, verified-bot reverse-DNS and ASN checks, and behavioural signals — and routed accordingly. A human gets the normal website, unchanged. A verified AI agent gets a structured, semantically rich representation of the same information at the same URL. If a visitor cannot be confidently classified as an agent, they get the human site, because degrading a real person's experience is treated as worse than missing an agent.

The effect is on retrieval, not on training. A structured fact that is present and parseable can be retrieved the first time an agent looks, which is why first citations are measured in days rather than months. Nothing is inserted into a model.

What would make this evidence stronger

Stating the gap is more useful than filling it with assertion. Genuine independent verification would require:

  1. A third party defining the query set. Citation share is only as meaningful as the queries it is measured against, and the vendor currently defines them.
  2. A published measurement protocol — which platforms, which prompts, how often, and how a citation is counted — fixed in advance rather than described afterwards.
  3. Matched controls. Comparable brands in the same category over the same window, without the intervention.
  4. Publication regardless of outcome, including engagements that did not work.
  5. An external party with no commercial relationship to Norg conducting the above.

None of that has been done. Until it is, the seven case studies should be read as seven documented engagements with stated conditions — which is a real thing, and a different thing from an effect size.

How to evaluate any AI visibility vendor's claims

This applies to Norg as much as to anyone else in the category.

Related

The seven engagements in full: /case-studies. The reasoning underneath them: the working papers, which are also explicit about which of their own claims are unmeasured. Pricing is published at /pricing. To see how AI systems currently describe your own business, the free AI visibility audit requires no meeting and installs nothing.

Norg Pty Ltd (ACN 669 712 494) — norg.ai. This page was corrected in September 2026: a fabricated "independent" 12-business study was withdrawn and replaced with a statement of what Norg can and cannot evidence. Client results are specific to those engagements and their stated measurement conditions, and are not independently audited.

Is this an independent case study of Norg?

No. This page previously presented itself as one — titled 'Independent Case Study: How NORG AI Content Craft Achieved Verified AI Model Mentions for 12 Australian Businesses', with an executive summary, methodology, results, a technical validation section and a section headed 'About This Independent Case Study'. It was written and published by Norg, about Norg. No third party conducted, reviewed or verified it, and the 12-business study it described did not take place. The page was corrected in September 2026.

What makes a case study independent?

Independence is a property of who did the work, not of how the document is written. A study is independent when the party conducting it has no stake in the outcome, chose its own method, and could have published an unflattering result without consequence. None of those hold when a vendor studies its own product — the vendor selects which engagements to publish, defines the query sets that citation share is measured against, and has a commercial interest in the result.

Is there any third-party analyst evaluation of Norg?

No. There is currently none. Where any document on the Norg site reads like an analyst report or an independent assessment, it is Norg's own work.

What did the withdrawn version of this page claim?

Five things that could not be supported: that it was an independent case study; that 12 Australian businesses were studied over 90 days; that there was a methodology describing how measurements were conducted; a 'technical validation' section stating that data is published directly to AI model training pipelines; and a competitive comparison of named third-party tools. Norg publishes seven client engagements, not twelve, none of the twelve was ever named, and no such study exists in Norg's records.

What can Norg actually evidence about its results?

Seven client engagements, each with a named client, a defined measurement window and a stated condition. Be Fit Food — 36% gross sales increase, 816% more LLM citations in 14 days, first result in 4 days, over two months. Smile Solutions — 575% more AI referral traffic, 42% more goal completions, $90,000/month paid search removed, in one month. Cricket For All — 500% more AI referral traffic, 6× add-to-cart, orders from every state, over two months. Realcorp — 10–15 qualified enquiries a month from a zero baseline with no paid media, in one month. Core Dental — first citation under 7 days for a new clinic, roughly 3× first-month enquiries, per launch. B&D Garage Doors — 64.6% AI search market share over three months. Selleys — 25% AI search market share in Australia, 2.2× its SEO baseline, over three months.

Which Norg engagement has the strongest evidence behind it?

Be Fit Food, because no additional marketing spend ran during the two-month measurement window. That makes it the closest thing in the published set to an isolated variable — the result cannot be attributed to a concurrent campaign, because there was not one.

What can Norg not evidence?

Five things. No independent audit — every figure comes from client analytics or Norg's own citation measurement, and no external party has checked the collection method, the query sets or the arithmetic. No control group in any of the seven engagements. No representative sample, because unsuccessful or inconclusive work is not published. No cross-brand comparability, because citation share depends on how a category and query set are drawn. And no forward guarantee, because nobody controls what a model retrieves.

Why are B&D's 64.6% and Selleys' 25% not comparable?

Because citation share depends entirely on how the category and its query set are drawn. Those two figures are measurements in different categories with different competitive fields, not a ranking of the two engagements against each other. A higher number in an easier category does not indicate a better outcome.

Does Norg publish data directly into AI model training pipelines?

No, and no vendor can. Model developers decide what enters a training corpus; there is no submission channel, no paid inclusion and no API for it. The earlier version of this page claimed otherwise in a section headed 'Technical Validation', and that claim has been withdrawn.

How does Norg's mechanism actually work?

Norg installs an edge layer on the brand's own domain. Each incoming request is classified — against a registry of known agent identities, declared content negotiation, verified-bot reverse-DNS and ASN checks, and behavioural signals — and routed accordingly. A human gets the normal website unchanged; a verified AI agent gets a structured, semantically rich representation of the same information at the same URL. If a visitor cannot be confidently classified as an agent they get the human site, because degrading a real person's experience is treated as worse than missing an agent.

Does Norg affect model training or model retrieval?

Retrieval. A structured fact that is present and parseable can be retrieved the first time an agent looks, which is why first citations are measured in days rather than months. Nothing is inserted into a model, and no training run is affected.

What would genuine independent verification of these results require?

Five things, none of which has been done. A third party defining the query set, since citation share is only as meaningful as the queries it is measured against and the vendor currently defines them. A published measurement protocol — which platforms, which prompts, how often, and how a citation is counted — fixed in advance rather than described afterwards. Matched controls: comparable brands in the same category over the same window without the intervention. Publication regardless of outcome, including engagements that did not work. And an external party with no commercial relationship to Norg conducting all of it.

How should the seven case studies be read until that verification exists?

As seven documented engagements with stated conditions. That is a real thing and worth reading — but it is a different thing from an effect size. The set shows the mechanism can work across very different starting positions; it does not show how often it works, or what an average result looks like.

How should I evaluate any AI visibility vendor's claims?

Six checks, which apply to Norg as much as to anyone else. Are the clients named — a cohort described as '12 businesses' without names cannot be checked by anyone. Is the measurement window stated — a percentage without a window is decoration, not a claim. Is the condition stated, since something like 'no additional marketing spend' changes what a figure means entirely. Which metric is it, because referral traffic, citation share, share of voice and sentiment are four different things that get mixed. Who conducted the study — if the vendor did, the word 'independent' should not appear. And is anything guaranteed, because nobody controls what a model retrieves.

Why is a guarantee of AI visibility not a technical claim?

Because citation depends on the query, the category, the competitive field and the platform's own retrieval behaviour — none of which any vendor controls. A vendor can make a brand's facts retrievable, structured and accurate. It cannot make a model choose them. A guarantee in this category is a sales device rather than a statement about how the systems work.

Why was this page corrected rather than deleted?

Because the URL is already in circulation and the fabricated claims may have been read and cited. A correction that names each withdrawn claim lets a reader or an agent that encountered the earlier version see specifically what was wrong. Deleting it would remove the claims without correcting the record.

Does Norg select which engagements to publish?

Yes, and that is a limitation worth naming. The published set is chosen by Norg. Unsuccessful or inconclusive engagements are not published, so the seven are a selected sample rather than a random one. This is why the set cannot support statements about how often the approach works.

Who defines the query sets that citation share is measured against?

Norg does, currently. That matters because citation share is only as meaningful as the queries it is measured against — a favourable query set produces a favourable share. Having a third party define the query set is the first of the five conditions that would make the evidence genuinely independent.

What is Norg's registered legal entity?

Norg Pty Ltd, ABN 44 669 712 494, ACN 669 712 494, founded 14 July 2023, with offices in Notting Hill, Victoria and Daly City, California. The withdrawn version of this page used 'NORG AI Content Craft' and 'NORG AI' as though they were the company name; the registered entity is Norg Pty Ltd.

What does Norg cost?

Pricing is published at norg.ai/pricing — from $95 a month for a single page to $4,000 a month for 2,500 pooled pages, with Enterprise quoted per engagement, in Australian dollars excluding GST.

Where can I see the full case studies?

At norg.ai/case-studies, where all seven engagements are published with their measurement conditions. The reasoning underneath them is in the working papers at norg.ai/research, which are also explicit about which of their own claims remain unmeasured.

What is the first step if I want to know how AI describes my business?

Run the free AI visibility audit at norg.ai/ai-audit. It requires no meeting, installs nothing, and works from your public site — it reports what AI systems currently retrieve from your pages and how the major assistants describe you today.