A funnel that sits underneath the source webinar rather than competing with it, built from the skills the deck already names. The webinar teaches SOURC-E and hands over three Claude Skills on day two, before the offer rather than after it. This funnel takes one of those three, puts it in front of the ad, and lets the output do the selling. A tool that runs on the buyer's own site beats any PDF, because it produces a result they did not have this morning and the result is the reason to buy. Everything here is a packaging and routing job: the skills exist, the zips exist in HawkOS Lite, the setup guide exists.
The hero skillStructure Audit. The deck makes the case itself: every other pillar multiplies against Structure, so nothing else counts until it is true. It is diagnostic, it is already one of the three handed over, and the raw-HTML demonstration is already built.
What stays paidThe judgement. Straight from the deck: the skills grade your site, a human still decides which fix matters most and whether the AI draft is any good, and that call is where the results live. That is why a free skill does not cannibalise anything.
The gateEmail and one qualifying question. No checkout, no call booking, no 8-field application. The cost of entry has to be lower than the $7 seat or this funnel has no reason to exist.
The real conversion eventNot the download. The install. A lead magnet that is never installed produces no score, and no score means no wedge. Measure activation, not opt-ins.
The ladderOne skill free, three at the webinar, the judgement on the call, HawkOS only with the team. Each rung gives away the thing below it, which is what makes the next rung believable.
Where it hands offThe webinar already ends on the strategy call, so this funnel does not need its own close. It needs to deliver people into the seat or the call already holding a score.
Fix before this launches: the landing page promises more than the webinar hands over. The deck is unambiguous. Day one says "three tools to run it yourself, handed over free". Day two's walking-away slide names them: Structure Audit, Original Data Finder and Credibility Packet Builder. Three Claude Skills. The closing stack slide names the same three again. Meanwhile the live landing page at webinar.studiohawk.com promises ten Claude skills, one or two per SOURC-E pillar, and the left-behind angle on the film day page says sixteen. So the page sells ten, the ad says sixteen, and the room receives three. That is an overstatement of more than three times on the single asset the offer is built on, discovered by every attendee at the exact moment you ask them to book a strategy call. Either write the page down to three, or hand over the other three named skills and write it to six. Do not run traffic to a ten-skill promise. Separately, the deck is branded source.studiohawk.com throughout while the ten-skill page is on webinar.studiohawk.com, so confirm which domain is current before any of this points anywhere.
The six skills, and the three that move
The asset
The deck names one Claude Skill per SOURC-E pillar as it teaches, six in total. Three of those six are actually handed over on day two. Every one exists today as a zip in HawkOS Lite, so nothing below needs building.
S · StructureStructure Audit ← handed over, and the free giveaway
O · OffsiteRAIDS Builder · taught, not handed over
U · UniquenessOriginal Data Finder ← handed over
R · RelevanceTopical Authority Mapper · taught, not handed over
C · CredibilityCredibility Packet Builder ← handed over
E · EvaluationMeasuring Results · taught, not handed over
Structure Audit is the right giveaway and the deck argues it for you: every other pillar multiplies against Structure, so nothing else you do counts until it is true. It is the only one of the three handovers that is purely diagnostic, it needs no strategy to produce a result, and the demonstration is already built, the rendered page next to the raw HTML the crawler actually gets. Give away one, not three. A three-pack sounds more generous and converts worse, because every extra skill is another install step and setup friction is the single biggest failure point of a tool lead magnet. Giving Structure Audit away also costs the webinar nothing it was keeping: attendees still receive it, plus two they did not have.
1. The ad
Stage one
Ad graphic text
The Tool We Use To Grade Client Sites. Yours Free.
Primary text
AI crawlers do not run JavaScript. If your content loads after the page does, the machine sees a logo, a menu and a footer, and decides you do not sell anything. We built a Claude skill that shows you exactly what the crawler gets. It is free, it installs in about fifteen minutes, and it runs on your site as often as you like. No call, no card.
60-second hook (film-day)
0-2s: "This is your website. And this is what an AI crawler actually sees."
2-15s: Same URL, same second. One is your site: copy, pricing, testimonials, the lot. The other is the raw HTML, which is a logo, a menu and a footer, because the content gets injected by JavaScript after load. To the machine, this business does not sell anything.
15-45s: GPTBot, ClaudeBot and PerplexityBot do not render JavaScript. We run this check on client sites before we quote them, and it is a Claude skill you can have. Claude is free, the skill is free, setup is about fifteen minutes with a guide written for people who do not code.
45-60s: No call, no card. Link below, put your email in and it is yours.
The promise has to be the tool and one specific, checkable fact, not a traffic outcome. Every competitor ad in this market promises more traffic. This one promises an object that arrives immediately, works without us, and settles a question the viewer cannot currently answer about their own site. The demonstration already exists in the day-two deck, so the film-day version is a screen recording, not a new build. Do not add urgency, there is none, and inventing it on a free tool reads as desperate.
2. The landing page and the gate
Stage two
Headline
See What An AI Crawler Actually Gets From Your Site.
Subhead
Structure Audit: the skill our team runs before we quote a client. Free, installs in about fifteen minutes, runs on your site as often as you like.
The gate (two fields only)
Email address
What is your annual revenue?
A) $5m+ B) $1m to $5m C) $250k to $1m D) Under $250k
What the page must show
A real screenshot of the skill's output, not a mockup. The single most common reason a tool lead magnet fails is that nobody believes a free thing does anything.
The install time, stated honestly, up front. Fifteen minutes.
That Claude is free and they do not need a paid plan.
One line on what it does not do, because naming the limit is what makes the rest credible.
Two fields, and revenue is the scored one so this funnel's leads sort into the same brackets as both topic tests and the numbers stay comparable across all three. Resist adding the "who runs marketing" field here: this is the top of the funnel and every extra field costs installs, which is the metric that matters. That question belongs on the call booking step, which only $1m+ answers ever see.
3. Delivery and setup
Stage three
Delivery email, subject lines to test
Your audit skill is ready
Here is the skill, plus the fifteen minute setup
The tool, the guide, and nothing else
Delivery email body
Here it is. Two links: the skill itself, and the setup guide.
The guide is written for people who do not write code. It takes about fifteen minutes and you only do it once. After that the skill lives inside Claude and you can run it on any site you own, as many times as you want.
When it finishes you will know whether an AI crawler can actually read your pages, where your heading structure stops making sense, and which of your commercial pages are invisible before any JavaScript runs. Reply to this email and tell me what it found, especially if it found nothing. I read these.
Harry
The reply request is the highest-value line in the funnel and costs nothing. It starts a real thread with a human before any sales step, it surfaces the people who actually installed, and the score they paste is better qualification data than any form field. Same mechanic as the manual touch in the call funnel confirmation, applied earlier.
4. The activation sequence
Stage four · where this funnel is won or lost
Day 1, if not installed
Subject: Stuck on the install?
The setup is the only hard part and it is about fifteen minutes. If you got stuck, reply and tell me where, and I will send you the fix.
Most people stall at the same step, so it is probably not you.
Day 3, if installed but not run
Subject: You installed it. Run it on one page.
It takes about two minutes. Point it at the page that makes you the most money, not your homepage.
Then reply and tell me what came back. If the answer is that your best commercial page is invisible to a crawler, you want to know that before you spend another dollar on content.
Day 5, if run
Subject: You just fixed the gate. There are five more.
Structure is the one everything else multiplies against, so nothing else you do counts until it is true. That is why we give this one away.
It is also only the S. There are five more letters, and they are where being readable turns into being chosen: who talks about you offsite, whether you publish anything nobody else has, whether you cover the whole question, whether the machine can verify you, and how you measure any of it.
That is the two days.
Track installs and runs, not downloads. A download is not a lead, it is a file transfer, and a funnel that reports opt-ins while nobody installs will look healthy right up until it produces no revenue. If the install rate is under about 30 percent the setup guide is the problem, not the offer, and no amount of ad spend fixes it.
5. The fork
Stage five · the score does the sorting
The skill hands you a segmentation you did not have to ask for: the score tells you how bad the problem is, and the revenue answer tells you who can afford to solve it. Route on both.
Door one: the $7 seatUnder $1m, or $1m+ and answering that they run marketing themselves. They have one score and no framework. The seat is SOURC-E end to end, plus Original Data Finder and Credibility Packet Builder on top of the skill they already hold, plus the slides, the prompt library and the six-question card. The free skill has already proved the toolkit is real, which is the hardest claim on that page to make believable cold.
Door two: the free strategy call$1m+ with somebody other than the founder running marketing. They have a score, a team and a budget, and they do not want to learn the framework, they want it run. Same offer the webinar closes on, so this door is not a new build. They arrive holding a diagnostic, which is exactly what the call is built to read.
This is the reason to build this funnel rather than write more angles for the two you have. Both existing funnels ask a stranger to self-report their situation on a form. This one watches what they do, gives them a number, and routes on evidence. A lead who installed a tool, ran it and replied with a score is a different animal to one who ticked a revenue box in an Instant Form. It also arrives at the call the same way a webinar attendee does, holding their own result, which is the state the call was designed around.
6. If they do not move
Stage six
The long nurture
They keep the skill. There is no takeaway, no expiry, no "your access ends Friday". They downloaded a tool that grades websites, which means they are the kind of person who wants to know where they stand, and that does not expire either.
Send them the thing the skill cannot do: one pillar of SOURC-E explained properly, once a fortnight, with an example. Six emails, six letters, then loop to the seat.
A free tool with no expiry is a permanent, useful presence on their machine that carries your name and re-runs whenever they think about search. That is worth more than a lead who unsubscribes in week two, so do not spend it on pressure. The people who never buy still run the audit, and some of them tell someone who does.
7. The HawkOS tier
The top of the ladder · not for sale as software
The skills are the consumer version. HawkOS is what the agency runs, and the deck already positions it correctly, as one of five moats and the only one nobody else has. Nothing on this page should ever offer HawkOS as a download, because the moment it is a product it stops being a reason to hire anyone.
The line from the deck, use it verbatim
We run it on our own operating system, the same one behind those results. It does in seconds what would take you a weekend, and it compounds. You'd be rebuilding it from scratch.
How the tiers differ, said plainly
The free skill grades one thing on one site, when you remember to run it.
The three webinar skills run the framework against your own site, when you remember to run them.
HawkOS runs the whole framework across a portfolio, on a schedule, with a specialist reading the output and deciding what to do about it.
The difference is not capability. It is whether it happens every week without you.
The other four moats, for the call and the nurture
Specialised: the skills grade your site, a human decides which fix matters most and whether the AI draft is any good. That call is where the results live.
Accountability: alone, this stalls the first busy week. With a team on it, it keeps moving, and in AI search the brand that keeps moving takes the citation.
Knowledge: we have already made the expensive mistakes across hundreds of campaigns. You would be making them on your own time.
Speed and cost: quicker than hiring in-house or building your own AI stack, with the SEO licences bundled in.
Note what this does to the giveaway objection. If somebody asks why you would hand over the tools, the answer is not a marketing line, it is the ladder: the tools are genuinely free and genuinely work, and the thing that produces results is the judgement and the momentum, neither of which fits in a zip. That is the deck's own argument, it is true, and it is the reason this funnel can be generous at the top without eroding anything below it. The honest limit to hold: never imply the free skill or the three webinar skills are a cut-down HawkOS. They are not the same product, and a buyer who works out they were sold a demo will not come back.
Wiring notes: The skill zips already exist in HawkOS Lite and the setup guide already exists, so nothing here needs building, only packaging and routing. Gate the download behind the two-field form and deliver by email rather than an instant redirect, because an email address you have not confirmed is worth very little and the delivery email is the first activation touch. Tag every lead with the angle that brought them in, the same way both topic tests do, so the three funnels stay comparable on one revenue question. Give away Structure Audit and nothing else: the moment Original Data Finder and Credibility Packet Builder are downloadable without the framework, the $7 seat is selling only the live session and the Q&A, and the toolkit stops being a reason to buy. Suppress this audience from the $7 webinar ads for the first fortnight so the free tool is not competing against a paid offer for the same person in the same week. The day-five email is the handoff and it should carry the seat, not the call, for anyone under $1m. Finally, the honest limit to state on the landing page: the skill tells you whether the machine can read your site, it does not tell you whether the machine will choose you, and it does not fix anything. That sentence is what makes the rest of the page believable, and it is also true.