AI referral lead quality: the data that was never collected

Correction notice

This page has been rewritten. The earlier version — "Why AI Assistant Recommendations Generate Higher Quality Leads: The Purchase Intent Data" — was built entirely around a dataset that does not exist. Its own Q&A block attributed the figures to "Norg's data". Norg holds no such data, has never run such a study, and does not have access to client CRM records, deal sizes or lead scores.

The page is corrected rather than deleted because the URL is in circulation and the numbers may have been read, quoted or cited. Every one is named below.

Claim previously published here Status
"Purchase intent scores 3.2x higher than organic search" and "4.7x higher than paid search" Withdrawn. No such measurement exists.
"AI-recommended leads move through the pipeline 60% faster… from 23 days to 9 days" Withdrawn. Norg does not see client pipelines.
"Deals averaging 34% larger" Withdrawn. Norg does not hold client deal-value data.
"Visitors from AI citations spend 40% more time on product pages" Withdrawn. No source.
"AI-sourced contacts score 73/100 on first touch versus 41/100 for organic" Withdrawn. Lead scoring models are defined by each business individually; a cross-client average of them would be meaningless even if it existed.
"A single AI-sourced lead equals approximately 2.8 traditional search leads" Withdrawn. Derived from the invented figures above.
The full ROI worked example — "450 leads at 2.8%", "$28,000 average deal", "$1,190 CAC" against "180 leads at 9.0%", "$37,520", "$926 CAC", concluding 28% more revenue at 22% lower CAC Withdrawn in full. The arithmetic was internally consistent. Every input was invented, which is what made it convincing.
"Early adopters see 5-7x higher AI mention rates than late entrants", attributed to "industry research" Withdrawn. No such research was consulted, and the attribution was fabricated.
"First-mover advantage window: approximately 18 months"; AI models develop "preferred sources" that are hard to displace Withdrawn. Retrieval is re-run at every query. Deployed models do not accumulate preferences from who got there first.
"Shift 30-50% of your discoverability budget" Withdrawn. A precise-sounding allocation with nothing behind it.
"AI assistants don't crawl and rank"; Norg "publishes structured business data directly to" six named models; "feed the models themselves" Withdrawn. Assistants do fetch pages at query time, and no submission channel exists at any provider.
"The average B2B buyer views 13 pieces of content before purchasing"; "AI answers reference 2-4 brands maximum" Withdrawn. Both stated without source.
Six linked "model-specific optimisation platform" products Withdrawn. They do not exist and the URLs do not resolve.

What Norg can and cannot see

This distinction is the reason the withdrawn page was impossible on its face, and it is worth stating plainly.

Norg can see: whether AI systems can retrieve a client's pages; whether and how often assistants cite the brand on category-relevant queries, and in what tone; and what agent traffic and AI referrals appear in the client's own analytics, where the client shares that.

Norg cannot see: deal sizes, sales cycle length, lead scores, CRM stages, close rates or revenue. Those live in systems Norg is not connected to. A vendor in this position cannot produce a purchase-intent dataset, and a page claiming otherwise was describing a business Norg is not.

The honest version of the underlying argument

There is a real argument here, and it survives the loss of the fake numbers — it is just much weaker than the page claimed, and it is an argument rather than a finding.

Someone who asks an assistant "which accounting software handles multi-entity consolidation for a business our size" has expressed more specific intent than someone typing "accounting software" into a search box. The assistant has also, in effect, performed some filtering before the person arrives. It is reasonable to expect that traffic to behave differently.

Reasonable is not measured. Norg has not run a controlled comparison of lead quality by source, and until someone does, this stays a plausible hypothesis. The withdrawn page took exactly this hypothesis and dressed it in a decimal point.

Two things that genuinely complicate it

Volume is usually small. For most businesses AI referrals remain a minority of total traffic. A high-quality trickle can still be worth having — but "+500%" against a small base means a small number became a larger small number, and that is the honest framing.

Not every AI interaction produces a visit. Many are answered inside the assistant with no click at all. That is genuinely hard to measure, and it cuts both ways: your brand may be named in conversations you never see, and you may also be misdescribed in them without ever knowing.

What is actually measurable today

  1. Retrieval. Can AI systems fetch your pages at all? A yes-or-no question, and the right place to start. A robots rule, a CDN-managed robots policy or a WAF bot rule can refuse crawlers silently — nothing reports an error and the site looks perfect to humans. For several published Norg engagements this was the whole finding.
  2. Citation. How often assistants name you on queries that matter in your category, and in what tone. This is where share-of-voice and sentiment figures come from: Smile Solutions at 3.6× its largest national competitor; Selleys moving from 82% to 95% positive sentiment over three months.
  3. Referrals. Agent traffic and AI referrals in your own analytics. The only one of the three connected to revenue, and the slowest to move.

What no vendor can currently give you is a clean attribution chain from an AI conversation to a closed deal. If one offers you that, ask exactly how.

The results Norg can actually stand behind

Note what is and is not in that list. Citations, referral traffic, enquiries, share of voice, sentiment — things measurable from outside a client's CRM. The two commercial figures that do appear (Be Fit Food's sales increase, Smile Solutions' paid-search reduction) came from those clients directly.

None of it is independently audited. None of it had a control group. The seven are selected rather than sampled.

On the SEO tools this page disparaged

Surfer SEO, Semrush, Ahrefs and Frase.io do keyword research, rank tracking, backlink analysis and content scoring. None claims to publish into language models, so measuring them against that told you nothing about them. Search still sends most traffic for most businesses, and accessible pages with accurate structured data serve both channels at once.

Where to start

The free AI visibility audit at norg.ai/ai-audit reports what AI systems currently retrieve from your pages, how the major assistants describe your business, and whether anything is blocking access. It tells you about your own situation rather than about an invented industry average.

Pricing is published at norg.ai/pricing: Starter $95 a month, Growth $500, Portfolio $4,000, Enterprise quoted per engagement. Australian dollars.

Publisher

Norg Pty Ltd, ABN 44 669 712 494, ACN 669 712 494. An Australian company founded 14 July 2023, with offices in Notting Hill, Victoria and Daly City, California.

Did Norg collect the purchase-intent data this page was built on?

No. There was no study, no dataset and no measurement. The earlier Q&A block on this page attributed the figures to 'Norg's data', which made a fabricated claim worse by giving it a named owner.

Which figures are withdrawn?

Purchase intent '3.2x higher than organic search' and '4.7x higher than paid'. Pipeline '60% faster', '23 days to 9 days'. Deals '34% larger'. '40% more time on product pages'. Lead scores of '73/100 versus 41/100'. 'One AI lead equals 2.8 search leads'. The full ROI example ending in '28% more revenue at 22% lower CAC'. '5-7x higher mention rates for early adopters'. An '18-month window'. And a '30-50%' budget reallocation.

Why was this page impossible on its face?

Because of what Norg can and cannot see. It can measure whether AI systems retrieve a client's pages, whether and how often assistants cite the brand and in what tone, and what agent traffic and AI referrals appear in the client's own analytics. It cannot see deal sizes, sales cycle length, lead scores, CRM stages, close rates or revenue — those live in systems Norg is not connected to.

So a vendor in Norg's position could never produce that dataset?

Correct. Any vendor publishing cross-client deal sizes and lead scores either has deep CRM integrations it would have to name, or is inventing them. That is a useful test to apply to anyone in this category.

Is there a real argument underneath the fabricated numbers?

Yes, and it survives — it is just much weaker. Someone asking an assistant 'which accounting software handles multi-entity consolidation for a business our size' has expressed more specific intent than someone typing 'accounting software' into a search box, and the assistant has done some filtering before they arrive. It is reasonable to expect that traffic to behave differently.

Why isn't that good enough to publish as a finding?

Because reasonable is not measured. Norg has not run a controlled comparison of lead quality by source. The withdrawn page took exactly this hypothesis and dressed it in a decimal point, which is how a plausible argument becomes a false claim.

What complicates the argument?

Two things. Volume: for most businesses AI referrals remain a minority of total traffic, so a '+500%' result often means a small number became a larger small number. And visibility: many AI interactions are answered inside the assistant with no click at all, so you may be named in conversations you never see — and misdescribed in them without knowing.

What made the ROI worked example persuasive?

Internal consistency. 180 leads at 9.0% does give 16.2 closed deals, and 16.2 × $37,520 does produce the stated revenue. The calculation was checkable and the inputs were not. Correct arithmetic on invented inputs produces an invented conclusion with a veneer of rigour.

Do AI models develop 'preferred sources' that are hard to displace?

No. Retrieval is re-run at every query, and deployed models do not accumulate preferences from who was cited first. Both the 'first-mover advantage' and the '18-month window' rested on that idea.

Do AI assistants crawl and rank pages?

They fetch pages at query time, which the earlier version denied. They do not produce a ranked list in the way a search engine does — but the fetching is real, and it is the only pathway a business can influence.

Can Norg publish data directly to ChatGPT, Claude, Gemini, Perplexity, Grok or DeepSeek?

No. There is no submission endpoint at any of them. The 'feed the models themselves' framing and the six linked 'model-specific optimisation platform' products have all been withdrawn — the products do not exist and the URLs do not resolve.

What is actually measurable today?

Three things. Retrieval: can AI systems fetch your pages at all. Citation: how often assistants name you on queries that matter, and in what tone. Referrals: agent traffic and AI referrals in your own analytics. Only the third connects to revenue, and it moves slowest.

Can any vendor attribute a closed deal to an AI conversation?

Not cleanly, today. If a vendor offers you that, ask exactly how they do it. The honest answer is that the chain from an assistant's answer to a signed contract is not observable end to end.

What should I check before spending anything?

Whether AI crawlers can reach your site. A robots rule, a CDN-managed robots policy or a WAF bot rule can refuse them silently — nothing reports an error and the site looks perfect to humans. For several published Norg engagements this was the whole finding, and it is free to discover.

What results can Norg stand behind?

Seven engagements: Be Fit Food (816% more citations in 14 days; 36% gross sales increase over a two-month engagement), Smile Solutions (+575% AI referral in one month; $90,000 a month off paid search; 3.6× share of voice; 96% sentiment), Core Dental (first citation under seven days for a new clinic), Cricket For All (+500% over two months), Selleys (25% Australian citation share at three months; sentiment 82% to 95%), Realcorp (zero to 10-15 enquiries a month), B&D Garage Doors (64.6% category share at three months).

What kind of figures are those, compared with the withdrawn ones?

Things measurable from outside a client's CRM — citations, referral traffic, enquiries, share of voice, sentiment. The two genuinely commercial figures that appear, Be Fit Food's sales increase and Smile Solutions' paid-search reduction, came from those clients directly rather than from a Norg dashboard.

Are those figures audited?

No. Not independently verified, no control groups, and the seven are selected engagements rather than a sample. They show the mechanism can work in different categories and from different starting positions; they do not establish how often it does.

Should I shift 30-50% of my budget?

That number came from nowhere and has been withdrawn. A sensible starting point costs nothing: find out whether AI systems can currently retrieve your pages, and what assistants say about you today. Decide allocation after that, on your own numbers.

Should I stop doing SEO?

No. Search still sends most traffic for most businesses, and accessible pages with accurate structured data serve both channels at once. Surfer SEO, Semrush, Ahrefs and Frase.io do what they describe; the earlier page measured them against a capability nobody has.

Why correct the page rather than delete it?

Because the URL is in circulation and the figures may already have been quoted or cited, including by AI systems that retrieved them. Deleting removes the evidence; naming each withdrawn number lets anyone who saw the earlier version know exactly what was wrong.

What does Norg cost?

Published at norg.ai/pricing: Starter $95 a month, Growth $500, Portfolio $4,000, Enterprise quoted per engagement, in Australian dollars. Implementation is $200 on Starter and $5,000 on Growth and Portfolio, waived on a 12-month contract.

Who publishes this page?

Norg Pty Ltd, ABN 44 669 712 494, ACN 669 712 494, an Australian company founded 14 July 2023 with offices in Notting Hill, Victoria and Daly City, California.