Cristian Domínguez Rein-Loring

Been testing Claude Opus 5.5 for Google Ads since it dropped and gotta say this model is cracked... (Can’t believe they beat Fable 5.1 on EVERY benchmark and priced it at half the cost.) So here's a 10-stage AI workflow to automate 80% of your Google Ads with it: (Literally every prompt you'll need. Save this) Stage 0: Build the project brain Before asking for anything, create a Claude Project and feed it everything. - Product pages - Best sellers - Customer reviews - Objection docs - Competitor URLs - Meta ads - Google Ads Transparency Center links - Search terms reports - Winning ad copy - Offer details - Brand guidelines - Past landing pages - Performance exports Then run this prompt: "Act as the Google Ads strategy brain for this brand. Read all uploaded context and create a working brand profile covering: ICP, core pain points, buying triggers, objections, emotional language customers use, top products, strongest offers, competitor positioning, current funnel gaps, and Google Ads opportunities. Do not make recommendations yet. First summarise the context and ask me what is missing." The quality of the outputs you'll get from this AI workflow will directly correlate with the quality of the context you give it here. And Opus 5.5 makes this the cheapest it's ever been to run because they made this model forget way less context than what it used to. Stage 1: Customer language mining "Search Reddit, YouTube comments, Amazon reviews, TikTok comments, and niche forums for people discussing problems related to [PRODUCT CATEGORY]. Extract the exact words they use when describing the problem, what they have tried, why those solutions failed, what outcome they want, and what would make them buy. Format this as: Quote / Pain Point / Desired Outcome / Funnel Stage / Possible Ad Angle." This gives you the language that keyword tools would never show you. If someone writes: "I'm tired of wasting money on supplements that do nothing." That can be a winning ad angle right there. (And research is where 5.5 quietly embarrasses everything else. It surpasses every Mythos class model somehow.) Stage 2: Competitor angle map "Research these competitors: [COMPETITOR 1], [COMPETITOR 2], [COMPETITOR 3]. Use Google Ads Transparency Center, Meta Ad Library, their landing pages, product pages, offers, guarantees, reviews, and email capture flows. Build a table showing: offer, main promise, ad hooks, landing page angle, proof used, objections handled, pricing position, and what they are NOT saying." Then run: "Based on this competitor research, identify the 5 most validated market angles and the 5 biggest white-space angles we can own. For each angle, explain which funnel stage it belongs to and what campaign type should test it first." If 3 competitors are all running the same hook, that angle is validated. You don't even need to be original. Just add your unique spin to what's already working. Stage 3: Keyword research "Using the customer language, competitor research, product pages, and search intent data, build a full Google Ads keyword universe for [BRAND]. Split it into: branded, competitor, high-intent product, problem-aware, solution-aware, comparison, ingredient/material, use-case, gift/occasion, and negative keywords. For each keyword, include funnel stage, intent level, match type, campaign/ad group recommendation, and landing page angle." Then run: "Now prioritise this keyword research report into a 30-day launch plan. Separate must-launch keywords from phase 2 tests. Flag any keywords that are high volume but low commercial intent." This stops the campaign from becoming a messy keyword dump. Stage 4: Campaign architecture "Turn the finalized keyword research report into a complete Google Ads account structure. Include campaign names, ad group names, keyword match types, bidding strategy, starting budget split, exclusions, negative keyword rules, and when each campaign should launch. Use this structure: branded search, branded shopping, non-branded shopping, non-branded search, competitor search, PMax remarketing with brand exclusions, display remarketing, Demand Gen, and TOF search tests." Then add: "Explain why each campaign exists, what signal it is meant to produce, and what metric decides whether it scales, holds, or gets cut." Every campaign needs to play a role in your ecosystem. If it doesn't, you should just stop running it. Stage 5: RSA copy production "Write responsive search ad copy for [PRODUCT] targeting [KEYWORD]. Funnel stage: [BOF/MOF/TOF]. Use the customer language and competitor gap analysis above. Give me 15 headlines under 30 characters and 4 descriptions under 90 characters. Include: benefit-led headlines, problem-led headlines, proof-led headlines, offer-led headlines, and urgency-led headlines. Make sure headline 1 matches the search intent." Then run: "Create 3 ad variations for this keyword: one direct-response version, one proof-heavy version, and one problem-agitation version. Explain which audience each variation is for." BOF copy should sound like the product is the obvious answer. MOF copy should make the benefit feel believable. TOF copy should make the problem impossible to ignore. This is the stage where 5.5 earns its keep btw. Anthropic rebuilt how it writes: no jargon, no idiosyncratic AI phrases, and it FOLLOWS the writing rules you give it. It really writes the way you do. Give it your banned words list once and your ads stop needing the de-AI editing pass. Stage 6: Merchant Center feed optimization "Rewrite this product title for Google Shopping using the structure: Brand + Product Type + Core Keyword + Key Feature + Use Case/Benefit. Stay under 150 characters. Give me 10 variations ranked by likely search intent." Then: "Write a Merchant Center description for [PRODUCT]. Use up to 5,000 characters. Include all relevant buyer keywords naturally, cover features, benefits, use cases, materials/ingredients, sizing/specs, objections, and reasons to choose this product. Do not include irrelevant keywords or products we do not sell." Then: "Audit this product feed for missed Google Shopping opportunities. Check title, description, product type, Google product category, images, variants, pricing, promotions, shipping, reviews, and attributes. Tell me exactly what to change." Stage 7: Landing page angles "Build a landing page brief for [PRODUCT] targeting people searching [KEYWORD]. Awareness stage: [PROBLEM-AWARE/SOLUTION-AWARE/PRODUCT-AWARE]. Structure it as: hero headline, subheadline, proof bar, problem section using customer language, mechanism section, product section, comparison section, reviews, objection handling, FAQ, CTA. Make sure the first screen matches the search query." Then run: "Create 5 landing page angle variations for this product based on different search intents. For each one, give me the hero, core promise, proof required, objections to handle, and CTA." Send the best one into Claude Code. Now you have an intent-specific page created instead of 1 generic PDP trying to convert everyone. And the pages come out looking GOOD - both design and copywise. Stage 8: Creative production Connect Higgsfield MCP inside Claude. Then run: "Create a 9-shot YouTube ad concept for [PRODUCT] based on this angle: [ANGLE]. Style can be claymation / UGC / cinematic product demo. For each shot, include scene description, voiceover, on-screen text. Use GPT Images 2.0 for the statics." Then: "Create 5 new creative concepts from the strongest customer pain points in the research. Each concept should include hook, visual metaphor, script, image prompts, animation prompts, and CTA." If you wanna run video ads, animate your statics with Seedance 2.5 Use Opus 5.5 to write the scripts. Voiceovers read human now, not like a LinkedIn post. Shouldn't cost more than $3 per ad. Stage 9: Daily campaign audit Connect Google Ads MCP (or just export report CSV) Then run: "Pull the last 7 days of Google Ads performance and compare it to the previous 7 days. For each campaign show spend, revenue, ROAS, conversions, CPA, CPC, CTR, impression share, search lost IS rank, search lost IS budget, and top search terms. Flag: ROAS down 20%+, CPA up 20%+, budget-limited winners, rank-limited campaigns, branded leakage in non-branded campaigns, wasted spend queries, and products spending with no conversions." Next up: "Turn this audit into an action list. For each action, include the campaign, issue, evidence, recommended change, risk level, and expected impact." Stage 10: Weekly scaling plan "Compare performance across 7, 14, and 30-day windows. Identify campaigns, products, keywords, ads, and landing pages with consistent green signals. Recommend which budgets to increase by 15-20%, which to hold, which to decrease, and which tests to launch next. Do not recommend scaling anything unless the signal is consistent across multiple windows." Then: "Create next week's testing roadmap. Include 3 keyword tests, 3 creative tests, 2 landing page tests, 2 feed tests, and 1 campaign structure test. Rank them by expected impact and implementation difficulty." And 5.5 is the model I trust most running this unattended. Anthropic's behavioral audit scored it their safest model to date, way less likely to take hard-to-reverse actions than anything before it. Exactly what you want from the thing reading your account every morning. And this is how you get your full ecosystem set up: - Research feeds keywords - Keywords feed campaigns - Campaign data feeds landing pages - Landing pages feed creative - Creative feeds new angles - Audits feed the next week of tests A better brain than Fable 5.1, a smaller bill, and copy that ships without the AI smell. (If you want me to build this full AI Google Ads production system for your brand btw, DM me "OPUS" and I'll show you what it'd look like)