Lines of work · 01
They ask an AILead generation
Inbound and outbound: what AI tells your customers, and the accounts you reach first.
Two runs at once. They find you: what the AI answers say about you, measured as ranges, with the wrong facts fixed where the answers get them. And you reach them: who buys written down, the lists built to that definition rather than bought, then outreach in two channels carrying proof from the last client.
How it runs, both ways at once
InboundThey find you
- 01
What the AI says now
The questions your customers ask, frozen before any work starts, and what each engine answers and cites.
- 02
A site the engines can read
Crawler access, bot blocking, and pages that only appear with JavaScript.
- 03
The same facts everywhere
Your site, Google Business Profile, LinkedIn and the directories the answers cite, all saying one thing.
- 04
On the lists they cite
The third-party lists and review sites the answers already draw on.
- 05
Measured again
At six weeks and at ninety days, against control questions the work never touched.
OutboundYou reach them
- 01
The ICP, written down
The accounts, the titles, and what they already pay to have the problem.
- 02
Lists built to it
Built to that definition rather than bought, then checked back against it.
- 03
Split with AI
The leads worth a person's time separated from the leads worth a sequence.
- 04
Hybrid outreach
Two channels at once, email and LinkedIn, each carrying proof from the last client.
- 05
Replies worked
Answered the day they land, and the accounts that went quiet worked again.
What you get
- What the AI answers say about you against your competitors, as ranges and never ranks, beside control questions the work never touches
- The wrong facts corrected at the source: your site, Google Business Profile, LinkedIn and the directories the answers cite
- A battlecard for your sales team: what AI says about you against each competitor
- The ICP written down: the accounts, the titles, and what they already pay for the problem
- The lists, built to that definition and split into high- and mid-value
- The sequences, in two channels, email and LinkedIn, with the social proof that lifts replies
What the AI answers say about you
Asked
“What expense software should a 60-person Australian company use?”
Answered
For a company that size, Competitor A and Competitor B are the usual picks: both handle corporate cards and sync with Xero. Your company also comes up, but it is mainly built for large enterprises1.
Named, out of 5 runs
- Competitor A
- 5 of 5
- Competitor B
- 4 of 5
- Your company
- 1 of 5
1Wrong. You sell to teams of 20 to 200. The engine cites a 2022 directory listing that no one updated.
How the answers are read
- The questions your buyers ask, frozen before any work starts, plus the ones that name you
- Five runs or more per question on each engine, logged out, from a Sydney location
- Four counts, never blended: mentioned, cited with a link, recommended, described accurately
- Every rate with its range, never a rank, and control questions the work leaves alone
On ChatGPT with search, Perplexity, Google AI Mode and Gemini.
What gets fixed, strongest evidence first
Let the engines read the siteMechanism
Crawler access, bot blocking at the CDN, content that only appears with JavaScript.
Make the facts agree everywhereMechanism
Pricing, integrations and who it’s for, the same on the site, LinkedIn and the review sites.
Get onto the lists the engines citeCorrelation
Brand mentions correlated 0.664 with visibility in Google's AI Overviews (Ahrefs, 75,000 brands).
Write a page only where nothing answersWeakest
In SAGEO Arena (KDD 2026), rewriting body text lowered visibility in a lab pipeline, so we correct the facts on the pages you have but don’t rewrite them for the engines.
Questions
- Can you get us named first in ChatGPT?
- Nobody can promise that. Answers change from run to run: two answers to the same prompt list the same brands less than 1 time in 100 (SparkToro and Gumshoe, January 2026). So we measure how often you are named, with a range, against control questions the work never touched, and fix the sources the answers draw on: the site the engines read, the facts they repeat and the lists they cite.
- Do we need llms.txt or schema markup?
- Google says its AI features need no special files or schema. Neither is sold here as a lever; the work is the site the engines can read, the same facts everywhere, and the third-party lists the answers already cite.
- What if it does not work?
- You will know, because everything is measured again the way it was measured first: the AI answers against control questions the work never touched, the replies against the baseline, and the pipeline leak by leak against the number each one was given before any work started.
The other lines of work
- 02 · They wait on a replyFunnel optimisationThe follow-up that leaks: the enquiry that waits, and the lead nobody chases.
- 03 · They hear your pitch from someone elseMeeting materials and pitchesA pitch that still works when someone else is telling it.
- 04 · They hear your pitch from someone elseSales trainingYour team trained on the questions that cost deals, the way debaters prepare.
What happens next
Talk to us about lead generation
Thirty minutes with our team