Prompts, concepts and activities. Every dark box has a Copy button — copy, paste into Claude / ChatGPT / Gemini, replace the [brackets] with your real numbers.
From 1 Sep, Google began moving Search campaigns to AI Max without asking. The rollout runs to 30 Sep, campaign by campaign. In Part 1 you audit what the machine did and decide. In Part 2 we climb the funnel from Google to Meta, and you build a Meta campaign on the right event.
Evidence first. Screenshots, change history, reports.
Fixed checks, in a fixed order. One finding per row.
One verdict. One reason. One re-check date. Written down.
Build the next thing on that decision.
You leave with: an audit sheet, and two runway calls.
You leave with: a paused-live Meta campaign and Ship Log layer 4.
AIDA is the oldest funnel in advertising: Attention → Interest → Desire → Action. It is a map, not a GPS. Real buyers loop back and forth. But it tells you which platform does which job.
India scale, for context: Instagram ad reach ≈ 481 million, Facebook ≈ 403 million, YouTube ≈ 500 million (DataReportal Digital 2026: India, Oct 2025 data). Google handles ≈ 98% of search referrals in India (StatCounter, Aug 2026). These are ad-reach figures, not monthly active users.
Google Search knows what a person wants right now, but little about who they are. Meta knows who the person is and what they do, but not what they want right now. That is why Meta starts high — and why targeting lets it reach down.
| AIDA step | Meta's targeting tool | What it does | Event to optimise |
|---|---|---|---|
| Attention | Broad / Advantage+ audience | Finds new people. You give signals; the machine picks who. | Reach · ThruPlay |
| Interest | Interests, behaviours, video viewers, page engagers | Points the machine at a habit. In 2026 these are hints, not fences. | Landing page view |
| Desire | Lookalikes seeded from your buyer list | Finds strangers who look like people who already paid. The seed quality decides everything. | Add to cart · Lead |
| Action | Custom audiences: site visitors, cart leavers, customer list | Retargets people who already know you. Exists only if the Pixel is collecting. | Purchase |
At ₹1L a month you will not run four campaigns, one per step. Next Sunday we build the consolidated two-layer stack (new people + people who know you). Today is the map.
A stranger stops on your Reel. No click yet.
Days later, they type your name into Google.
Your Week 2 brand campaign takes the click.
Visitors and buyers become Meta audiences.
| Carries over from Google | Changes on Meta |
|---|---|
| A structure you can read: campaign → ad set → ad | You create the want; you do not harvest it |
| The success metric, chosen before you spend | The lever is the optimisation event, not the keyword |
| The volume math: can your conversions feed the algorithm? | Creative does the targeting work (next Sunday) |
| Audit the machine — Advantage+ defaults get the six checks too | Last-click is the wrong ruler |
| C | Context | Budget per month, AOV, net margin, last month's purchases or leads, your override rule, both evidence packs. |
| R | Role | The expert hat: Google Ads auditor · YouTube media planner · Meta media buyer · front-end developer. |
| I | Intent | One job at a time. Audit first. Verdict second. Build third. |
| S | Specifics | The six checks in order · ~50 events per ad set per week · the graduation trigger · your exact CTA selector. |
| P | Parameters | Show the math · name the lever · flag what you cannot see · end with a checklist. |
CONTEXT Indian Google Ads account. Monthly budget ₹[budget]. My override rule: CPA ceiling ₹[X]; act if CPA runs above 1.3× the ceiling for 3 days. BEFORE pack ([date, e.g. 31 Aug]): [paste settings + search terms + assets] AFTER pack ([date, e.g. today]): [paste settings + search terms + assets + change history] ROLE Act as a senior Google Ads auditor who is sceptical of platform claims. INTENT Turn these two packs into one audit sheet per campaign. SPECIFICS 1. List every change between the packs. For each one say who made it: me, an auto-applied recommendation, or the AI Max upgrade. Give the date. 2. Run six checks, in this order: change history · search terms beyond my keywords · asset honesty (any claim I cannot prove?) · landing pages · brand and geo controls · CPA vs my rule. 3. If the upgrade is less than 2 weeks old, do not judge CPA — say "collect". PARAMETERS Per campaign give ONE verdict: keep / steer / opt out, with the exact lever (negative list, brand exclusion, URL exclusion, text customisation off, search-term matching off) and a re-audit date. Finish with a list of everything you cannot tell from screenshots.
CONTEXT Indian [D2C / EdTech / services] brand. Total paid budget ₹[budget]/month. Blended CAC target ₹[X]. My intent map from Week 2: [paste] ROLE Act as a YouTube media planner who has to defend the spend to a CFO. INTENT Tell me whether YouTube earns a seat at this budget — yes, not yet, or retargeting only. SPECIFICS Goal: [awareness / consideration / action]. If yes: the format, the targeting (start from my intent map, not broad demographics), the video length, and the view → remarketing audience bridge I must build. PARAMETERS Name the ONE metric to judge it on at this funnel stage, and the metric I must never judge it on. Show what % of budget I can give a channel I cannot last-click for 4–8 weeks and still hit blended CAC. State the minimum monthly budget below which YouTube is not worth running.
CONTEXT Indian [category] brand, ₹[budget]/month paid budget, AOV ₹[Y]. Monthly searches for my category keywords: [N, from Keyword Planner]. Monthly searches for my brand name: [N, from Search Console]. Channels I run today: [list with monthly spend]. ROLE Act as a full-funnel media strategist for Indian brands. INTENT Map my channels onto the AIDA funnel and find my biggest gap. SPECIFICS 1. Place each channel I run on Attention / Interest / Desire / Action. 2. Tell me which step nobody is paying for right now. 3. Is my category search volume big enough to grow on Google Search alone? Show the math: searches × realistic click share × conversion rate. 4. If not, what is the cheapest way to create demand on Meta at my budget? PARAMETERS One table, one gap, one recommendation. Tell me which number in Search Console will move first if Meta is working. Flag every assumption.
CONTEXT Indian [D2C / lead-gen] brand. Meta budget ₹[budget]/month. Goal: [sales / leads]. Offer: [X]. AOV ₹[Y]. Net margin ₹[Z]. Last month's total purchases or leads, all channels: [N]. ROLE Act as a Meta media buyer who has managed thin Indian budgets. INTENT Choose my campaign objective AND my exact optimisation event. SPECIFICS 1. Name the cheaper event that will mislead me, and why. 2. Can my volume feed Purchase at ~50 events per ad set per week? Show the math: budget ÷ expected CAC ÷ 4.3 weeks ÷ number of ad sets. 3. If it cannot: the proxy event to use for now, and the graduation trigger — the exact weekly number at which I switch to Purchase. 4. How many ad sets should I run at this budget? Default to one. PARAMETERS State week-1 success as ONE KPI (not cheapest CPM). End with the Pixel events I must see in Events Manager before I spend a rupee.
CONTEXT Indian [category] brand. Meta budget ₹[budget]/month. AOV ₹[Y]. My Pixel shows, last 7 days: [visitors] visitors, [CTA events] CTA events, [buyers] buyers. Customer list with consent: [N] contacts. ROLE Act as a Meta audience strategist working under Advantage+ in 2026. INTENT Give me a targeting plan that covers the funnel without splitting my budget into ad sets that are too small to learn. SPECIFICS 1. For Attention, Interest, Desire and Action: which Meta targeting tool fits (broad, interest hints, lookalikes, custom audiences) and whether I can afford that step yet. 2. Which inputs are only suggestions the machine may go beyond, and which are hard limits (location, minimum age, language, exclusions)? 3. Which audiences are too small to use today, and the visitor or buyer count at which each becomes usable. 4. Who must I exclude so I do not pay to reach existing buyers? PARAMETERS Maximum two ad sets. Show the ~50 events per week math for each. Flag anything that needs customer consent under India's DPDP rules.
CONTEXT My static site is on Netlify at [URL]. My index.html head is: [paste your <head> … </head>] My Meta Pixel ID is [ID]. My main CTA is [button selector or form id]. ROLE Act as a front-end developer who sets up tracking for small Indian brands. INTENT Install the Meta Pixel and fire one event when my CTA is used. SPECIFICS 1. Give me the exact Pixel base code and where it goes in my head. 2. Add one event on my CTA click or form submit — the standard Lead event if it fits, otherwise a named custom event. 3. Keep my existing GA4 tag untouched. PARAMETERS Return the full edited head, ready to paste. Then a 3-step check using Events Manager → Test Events. Tell me what to re-drag to Netlify, and what an ad blocker will hide while I test.
PageView and your CTA event. GA4, Pixel and (next Sunday) CAPI are three views of one event — they will never match exactly.Copy the template and paste it into cell A1 of a Google Sheet. It lands in columns. Six rows, one finding each, one verdict at the bottom.
Campaign: [name] Upgrade date: [from change history] Check Finding Who made the change (me / auto / upgrade) Lever 1. Change history 2. Search terms beyond keywords negative, list level 3. Assets — true, yours, provable? text customisation off 4. Landing pages URL exclusion 5. Brand + geo controls brand exclusion · location 6. CPA vs my rule collect if under 2 weeks VERDICT (keep / steer / opt out) Reason: Re-audit date:
| Verdict | When | What you actually do |
|---|---|---|
| Keep | Terms clean, assets honest, CPA inside the rule, volume up | Leave it. Audit weekly. Judge it 2 weeks after its upgrade date. |
| Steer | Good volume, but waste or bad assets | Negatives at list level, brand exclusions, URL exclusions, text customisation off. Re-audit in 7 days. |
| Opt out / rebuild | Waste you cannot steer — or you need a clean read of intent | AI Max off; or rebuild as exact / phrase. Keep the negative list. |
| Worked at ₹1L / month · ₹799 AOV | Purchases / week | Verdict |
|---|---|---|
| Meta budget ~₹60K at ~₹320 CAC → ~185 purchases / month | ~43 (one ad set) | Borderline. You clear it only if you do not split. |
| Same budget across 3 ad sets | ~14 each | Permanent learning. Erratic and expensive. |
| Founder at ~100–120 purchases / month in total | ~25 | Below 50. Purchase starves. Use a proxy event with a written graduation trigger. |
| The Add-to-Cart trap | Optimise: Add-to-Cart | Optimise: Purchase |
|---|---|---|
| CPM | ₹180 (cheap!) | ₹320 |
| Headline cost | ₹40 / add-to-cart | ₹320 / purchase |
| Add-to-cart → purchase | ~6% | — |
| True cost per buyer | ~₹670 | ₹320 |
| vs net margin ₹275 | loses ₹395 per buyer | loses ₹45 on order 1 — pays back on order 2 |
Illustrative bands for teaching (D2C conversion CPMs ₹150–400, 2026). Your own account's numbers outrank every band here within three weeks.
PageView and your CTA event fire for 3 days in Events Manager.Next Sunday — Build the Machine: audiences, Advantage+ defaults you choose, CAPI, creative teardown.