The PetitionFunnel

The template advice says: target your city, exclude every other country, and put your service and your suburb in the headline. Applied literally to Sherrod Sports Visas, that advice would switch off your entire market on day one. Your clients are not in Los Angeles, Charlotte or Washington. They are in São Paulo, Lagos, Tbilisi, Manila and Belgrade, searching in the middle of their night for a way to fight, play or coach in the United States.

So before a single keyword goes in, hold four facts in mind. Everything in the rest of this book is derived from them.

What "winning" means here

Clicks are vanity. Form fills are closer, but still not money — a free visa evaluation from someone with no ranking, no national team history and no US itinerary costs you consultation time and returns nothing. The only number that governs decisions in this book is revenue per dollar of ad spend, by matter type.

That matters more than usual because your matters are not fungible. A DIY course sale, a P-1S support-staff petition, an O-1A for a world champion, and a federal litigation matter are separated by orders of magnitude. An account optimised for "leads" will happily buy you a thousand of the cheapest, least qualified enquiries. Chapter fourteen fixes that by feeding real matter values back into Google.

A Georgian wrestler types p-1 visa for wrestlers. If your ad says "US Immigration Attorneys — Full Service", he does not recognise himself in it and scrolls on. If the ad says "P-1 Visa For Wrestlers" and the page he lands on says "P-1 Visas For Wrestlers", he has been told three times in eight seconds that he is in exactly the right place.

That is the whole trick. It sounds too simple to be worth a chapter, and it is the difference between a two percent and a twenty percent conversion rate. Google also rewards it directly: relevance between keyword, ad and landing page is a component of Quality Score, which sets what you pay per click.

Ask any model for a keyword list and you will get a plausible, confident, partly fictional set of phrases — some with no search volume at all. Search volume is empirical. It comes from one place: Google Keyword Planner, which is free inside any Google Ads account.

Pulling the raw list

Sorting winners from losers

Go down the export and classify every row. The test is not "is this related to visas" — it is "would the person typing this pay for a petition?" Here is how the main patterns break for you.

The rule that keeps you honest

Never build a cell that Keyword Planner says has no volume. A page for "O-2 visa for competitive fishing" will sit there getting three impressions a month and teach you nothing. Where volume is thin, roll the cell up: one "combat sports" ad group instead of separate MMA, muay thai, kickboxing and grappling groups, until the data justifies splitting them.

One clarification worth stating plainly, because the word is misleading: a "keyword" is not a word. It is the whole search someone typed. where can i get a p1 visa as a rugby player is one keyword. That is why phrase match wins — it catches the thousand ways people phrase the same underlying request without opening the door to boxing gyms and salary questions.

Why one keyword per group

Because it forces alignment. The moment an ad group holds fifty keywords, its ads must be written vaguely enough to cover all fifty — and vague ads are the generic bottom row of Plate 01. With one keyword, the ad can say precisely what the searcher typed, and the landing page can too.

The cost of this approach used to be labour. A hundred cells means a hundred ad groups, several hundred ads and a hundred pages. That is the exact cost Claude Code removes.

Why several ads inside each group

You are not looking for a good ad. You are looking for an outlier — the one variant that dramatically outperforms and that you then run for years. You find outliers by testing many, not by writing one carefully.

One caveat that matters more for you than for a plumber: you need enough volume for a test to mean anything. Some of your cells will get a few hundred impressions a month. Three variants is plenty there. Reserve the ten-variant tests for your highest-volume cells — combat sports, soccer, and the category terms like "sports immigration lawyer".

The pattern is always the same: put a markdown file in the project that encodes the standard you want, then write a prompt that tells Claude to read it and act. The prompt is short. The file does the work. This is why the output is specific to sports immigration rather than generic marketing filler.

Set your project up with these files before you build anything.

Notice three things about that prompt. It asks for a research summary first, so you catch misunderstandings before they get baked into a hundred ad groups. It scopes the task tightly. And it asks for approval before anything goes live. Keep all three habits.

The settings that matter, and what to set them to

Every Google Ads guide tells you to target your city and exclude the rest of the world. That advice exists because most advertisers are local service businesses. You are the exception that proves the rule: your revenue comes almost entirely from people who are not in the United States and want to be.

So the exercise is not exclusion. It is ranking the world by athlete export and building a tiered geographic bid structure.

Every time someone searches, an auction runs. Roughly: higher bids and higher quality get higher positions, and higher positions cost more per click. The lever most advertisers reach for is the bid. The lever that actually compounds is quality.

Where to start, and where to end up

Start on Maximise conversions. Google needs conversion volume before its models are useful, and a brand new account has none. Once conversions are flowing reliably, move toward target CPA, and finally to target ROAS once real matter values are being uploaded back into Google — which is Chapter Fourteen and is where this account becomes genuinely dangerous to your competitors.

The three cost levers, in order of leverage

Reviewing what comes back

Do not skip this because the volume is large. Spot-check five ads per campaign against three questions:

• Is headline one the literal search term, and is it the only pinned asset?
• Does any description promise, guarantee or imply a visa outcome? If so, fix brand-voice.md, not just the ad — otherwise it recurs in every future batch.
• Are the sitelinks pointing at pages that actually exist and are relevant to that ad group?

When you find a systematic problem, the fix belongs in the reference file. That is the difference between using Claude Code as a text generator and using it as a system.

A negative keyword stops your ad showing for searches containing that term. Building this list well is the highest-return unglamorous work in the account. Every blocked junk click is money that stays in your budget for a real petition enquiry.

Group them into themes so you can reason about them later, and share one list across all campaigns.

Your buyer has one dominant question, and it is not "what is a P-1A". It is: have these people got someone in my situation into America before? A Georgian wrestler, a Brazilian jiu-jitsu black belt, a Nigerian sprinter — each is scanning for evidence that their specific case is familiar to you. Your approvals archive and your named client wins are the most valuable conversion assets you own, and most firms bury them.

The structure of one page

Two more things, briefly. Split test everything — my opinion, your instinct and your colleague's certainty are all guesses until traffic decides. And note that until you deploy (Chapter Fifteen), these pages run only on your own machine; the ads' final URLs stay on a placeholder until the pages are actually live.

Cold traffic does not know you. Warm traffic has read your approvals page, compared O-1 against P-1, and come back. The second group converts at a multiple of the first. To advertise to them you need Google to know who they are — which means a tag on your pages and an audience built from it.

This matters more in your business than in almost any other. Nobody engages an immigration firm on the first click. They research, they ask their agent, they talk to a training partner who did it last year, and then they come back — sometimes weeks later.

Setting it up

• Create setup-conversion-tracking.md with your conversion actions defined — evaluation form submitted, WhatsApp click, pricing page view, DIY purchase — and each one's value.
• Run the tracking prompt. Claude places the tag and wires the hidden fields.
• Verify with Google Tag Assistant: connect to the page, complete a test submission, then paste the Tag Assistant output back into Claude and ask it to confirm the tag and conversion events are firing correctly. Reading that output yourself is unnecessary work.
• Build the audiences: all visitors, pricing-page visitors, form-starters who did not finish, and DIY purchasers. Different messages for each.

Which remarketing to run

Search remarketing (RLSA) — yes. Same search ads, shown to people who already visited. Higher cost per click, much higher conversion rate. This is your best remarketing channel by a distance.

Display — try it, cautiously, warm only. Banner ads to people who already know you can work. Banner ads to cold audiences, and Performance Max, Demand Gen and video for a considered legal purchase, generally buy accidental taps rather than clients. If you test them, cap them hard and judge on signed matters, not impressions.

Traditional optimisation is slow, technical and dull: which campaigns to pause, which to scale, which search terms to block, whether the bid strategy is still right. It is exactly the kind of work a busy practice never gets to and an agency charges thousands a month for.

Build a dashboard once, then read it every Monday and let Claude execute the fixes.

The Monday routine

Open the dashboard. Read the recommendations. Then run these three prompts in order — they take about ten minutes and replace most of what a retained agency does.

Right now Google has no idea that the enquiry from a ranked featherweight is worth twenty times the enquiry from an amateur with no US bouts. Left to guess, it treats every conversion as equal and cheerfully buys you a flood of the cheap ones. The fix is a data loop from click to signed matter and back.

Step five is the one people skip, and it is the one that changes the account. Until the fee data goes back into Google, all your careful revenue tracking is a report you read — useful, but it does not steer the bidding. Once uploaded, target ROAS bidding starts hunting for people who resemble your paying clients.

The implementation differs for every business because it depends on your CRM and billing stack, so this is a conversation with Claude rather than a fixed recipe. Describe your stack and work through it link by link.

A skill is a saved workflow you invoke by name. Instead of re-explaining your voice, your structure and your compliance limits every time, you type a slash command and a short line of context. This is what turns a one-off build into an operating system.

After that, expanding coverage is a sentence. /generate-ads five more variants for the O-1A esports ad group. /new-landing-page for p-1 visa for rugby players. A new sport vertical goes from a week of work to an afternoon.

A sane operating rhythm

• Monday, 15 minutes. Dashboard, then /audit-negatives, then approve or reject.
• Fortnightly. /generate-ads for any ad group down to two live variants. Retire clear losers.
• Monthly. /geo-review and the full account audit. Promote Tier 2 winners.
• Quarterly. Re-run keyword research. Sports rise and fall; new leagues and promotions appear. Add matrix cells that now have volume.