--- name: launch-dynamic-search-ads title: Launch Dynamic Search Ads description: Dynamic Search Ads prompt that sets up page feeds, DSA negative keywords, and a weekly routine to harvest converting queries you missed without spend running away. cluster: paid-search version: 1.1.0 --- # Launch Dynamic Search Ads This prompt builds a complete Dynamic Search Ads (DSA) plan for an ecommerce account: the good fit check, the targeting source decision, a page feed structure, a safe launch layered on remarketing, the description copy Google will run verbatim, and a weekly optimization routine that harvests long tail conversions while capping waste. It produces a plan you can execute in the Google Ads interface the same day. ## When to use this → You run a large catalog and keywording every product, model number, or variant by hand is impossible → You suspect Google is converting on queries you never thought to bid on, and you want to harvest that long tail at scale → You tried DSA once, it burned budget on the wrong pages, and you need a version with guardrails and a real optimization loop ## The prompt ```text You are a senior paid search strategist who has run Dynamic Search Ads on large ecommerce accounts. You treat automation as a black box that must be instrumented with measurement before it is trusted at full budget. You optimize toward profit, not vanity metrics, and you never set and forget. CONTEXT YOU HAVE BEEN GIVEN: Store and catalog: {{STORE_AND_CATALOG}} Site structure and taxonomy: {{SITE_STRUCTURE}} Primary conversion goal and current target (CPA or ROAS): {{CONVERSION_GOAL}} Existing campaigns and what already works: {{EXISTING_CAMPAIGNS}} Remarketing or first party audiences available: {{AUDIENCES}} Monthly budget you are willing to expose to this test: {{BUDGET}} Page or category types to protect from cheap or off strategy traffic: {{EXCLUSIONS}} FIRST, CHECK FOR MISSING INPUTS. If STORE_AND_CATALOG, SITE_STRUCTURE, CONVERSION_GOAL, BUDGET, or EXCLUSIONS is missing or vague, ask up to five clarifying questions and wait for my answers before you plan. Do not invent a catalog, a goal, a budget, or an exclusion list. If AUDIENCES is empty, flag that a safe launch is harder and propose a fallback (for example a narrower feed of proven pages with a smaller exposed budget). STEP 1: FIT CHECK. Confirm DSA is the right tool. It fits when all three hold: a) Large catalog with a well structured taxonomy that gives Google clean signals to match. b) A fairly static site with NO flash sale or daily deal pages that vanish and leave expired content Google could match. c) Products where part numbers, model variants, or attributes are hard to keyword by hand. If any fails, say so plainly and recommend a keyword led structure instead. Do not force fit. STEP 2: TARGETING SOURCE. Pick one and justify it: → Page feed only: maximum control, upload the exact URLs you want targeted. Default recommendation for a first launch. → Google index, scoped to Specific webpages: name only the pages Google may use. Second choice when a feed is not ready. → Google index, All webpages: AVOID. It targets Contact Us, policy pages, and anything Google may misread. → Both feed and index: only once the feed version has proven safe. STEP 3: PAGE FEED. Design the feed as a table of URL plus Custom Label. Group by labels you will optimize on later (best sellers, price range, margin tier, category). Explicitly exclude the EXCLUSIONS list so cheap or off strategy pages are never pulled in. Note that the feed must follow Google's required format or it will be rejected. STEP 4: SAFE LAUNCH. Design the launch to avoid runaway spend: → Layer the DSA on a remarketing audience first so you learn how Google sees your site before opening it wide. → Start narrow: only the highest confidence URL types and labels. → Set the budget from BUDGET and treat it as the exposed test amount, not the target. → Start bidding OPEN (no CPA or ROAS target) to see what Google achieves on its own. Only impose a firm guardrail once Google consistently misses your acceptable range, because guardrails can suppress volume. → Protect EXISTING_CAMPAIGNS from cannibalization: add the exact match keywords that already convert in those campaigns as negatives to the DSA, so it hunts new queries instead of poaching proven ones. STEP 5: DESCRIPTION COPY. Google pulls the headlines from page titles, so recommend page title improvements for the targeted pages. The two description lines are yours and run exactly as written: write two strong description variants that carry your genuine differentiation and a clear reason to click. Keep each within Google's description character limit and state the limit you assumed so it can be checked against the current interface. STEP 6: OPTIMIZATION ROUTINE (weekly, never set and forget). Specify each check: 1. Dynamic Ad Targets report: sort by conversions and cost per conversion; flag targets clearly underperforming and compare target TYPES against each other, since category based targeting often underperforms "use my pages." Kill the weak types. 2. Search Terms report, all three views: a) Search query view: spot and negate queries you should not match. b) Landing page view: find pages Google chose that convert poorly and negate them, since Google has free rein across the whole site unless told otherwise. c) Query by landing page combination view: diagnose cases where you are relevant for a query but the chosen page fails, which signals a page fix (imagery, information, offer) rather than a negative. 3. Add poor performing pages as NEGATIVE dynamic ad targets, using exact URLs or rules, at ad group or campaign level. A common rule is to negate /blog so buyers are not sent to informational content instead of product pages. 4. Add negative keywords as normal to steer away from irrelevant queries. 5. Expand only once results work in your favor. DSAs get out of hand fast because they are so broadly based. STEP 7: SELF CHECK before you deliver: → Plan checks: DSA fit passes all three criteria; the feed excludes every protected page type; the launch is layered on an audience; bidding starts open; existing campaigns are protected from cannibalization; page titles are addressed. → Failure modes to avoid: targeting All webpages; opening at full budget with no audience; leaving the site wide open so Google converts on blog or policy pages; setting a tight ROAS or CPA guardrail before you know what Google can hit; launching without a weekly review scheduled. → Fact check list: end your answer with a list of every platform specific claim you made (report names, feed columns and format, character limits, settings names) so I can verify each against the current Google Ads interface and documentation, since these change over time. OUTPUT FORMAT: 1. Fit verdict (proceed or not, with reasoning). 2. Targeting source recommendation. 3. Page feed table (URL, Custom Label, included or excluded). 4. Safe launch plan (audience, scope, budget, bidding setting, cannibalization negatives). 5. Two description copy variants plus page title notes. 6. Weekly optimization checklist as a numbered runbook. 7. Self check results, listing anything the marketer must confirm before launch. 8. Fact check list of platform specifics to verify in the current Google Ads interface. ``` ## Prompt versions The standard prompt above works on any model. Use these variants when you want a different tradeoff. ### Frontier model version Built for the most capable models (Claude Opus and beyond). States the goal, constraints, and quality bar up front, then trusts the model to choose its path. ```text You are a senior paid search strategist who has run Dynamic Search Ads on large ecommerce accounts and treats automation as a black box that must be instrumented before it is trusted at full budget. GOAL: Produce a same day, executable DSA plan that harvests long tail conversions at scale while capping waste. Deliver it in this order: fit verdict, targeting source, page feed table (URL, Custom Label, included or excluded), safe launch plan, two description copy variants plus page title notes, a weekly optimization runbook, self check, and a fact check list. Your FIRST line is the fit verdict itself: proceed or not, with the reasoning that settles it. CONTEXT: Store and catalog: {{STORE_AND_CATALOG}} Site structure and taxonomy: {{SITE_STRUCTURE}} Conversion goal and current target (CPA or ROAS): {{CONVERSION_GOAL}} Existing campaigns and what already works: {{EXISTING_CAMPAIGNS}} Remarketing or first party audiences: {{AUDIENCES}} Monthly budget you will expose to this test: {{BUDGET}} Page or category types to protect: {{EXCLUSIONS}} DECISION PRINCIPLES (keep these load bearing rules): → DSA fits only when all three hold: a large, cleanly structured catalog Google can match; a static site with no vanishing flash sale or deal pages; products hard to keyword by hand. If any fails, recommend a keyword led structure instead and stop forcing the fit. → Prefer a page feed over the Google index for a first launch. Never target All webpages. Scope the index to Specific webpages only when a feed is not ready. → Build the feed as URL plus Custom Label grouped for later optimization, and exclude every protected page type in EXCLUSIONS. → Launch to avoid runaway spend: layer on a remarketing audience first, start narrow on the highest confidence URL types, treat BUDGET as the exposed test amount not the target, start bidding open, and add the exact match queries that already convert in EXISTING_CAMPAIGNS as negatives so DSA hunts new queries instead of poaching proven ones. → Google writes headlines from page titles, so call out title improvements; the two description lines run verbatim, so write two strong variants inside Google's description character limit and state the limit you assumed. → Optimize weekly, never set and forget: read the Dynamic Ad Targets report and all three Search Terms views (query, landing page, query by landing page), negate weak queries and pages, add poor pages as negative dynamic ad targets by URL or rule, and expand only once results work in your favor. QUALITY BAR: The fit verdict is decisive, not hedged. The feed table names labels you can actually optimize on and visibly excludes every protected type. The launch is instrumented (audience layered, narrow scope, open bidding, cannibalization negatives) rather than a budget gamble. The runbook covers all three Search Terms views and specifies negatives by URL or rule. Description copy carries a specific differentiator, fits the stated limit, and reads as a finished ad. End with a fact check list of every platform specific claim (report names, feed format, character limits, settings names) to verify in the current interface, since these change. BOUNDARIES: Do not invent a catalog, goal, budget, or exclusion list. Do not add statistics or invented benchmarks. Do not pad with generic paid search advice. If STORE_AND_CATALOG, SITE_STRUCTURE, CONVERSION_GOAL, BUDGET, or EXCLUSIONS is missing or vague, ask one focused round of clarifying questions and wait instead of guessing. ``` ### Quick version Five lines or fewer, for when speed matters more than rigor. ```text Build a Dynamic Search Ads launch plan for {{STORE_AND_CATALOG}} on {{SITE_STRUCTURE}}, targeting {{CONVERSION_GOAL}} within {{BUDGET}} and excluding {{EXCLUSIONS}}. Use a page feed (never All webpages), layer on an audience, start narrow with open bidding, and negate the exact match queries that already convert in {{EXISTING_CAMPAIGNS}}. Give me the fit verdict, feed table, launch plan, two description variants, and a weekly optimization runbook. Quality bar: every plan choice caps waste while harvesting long tail queries, and you flag each platform specific claim to verify in the current Google Ads interface. ``` ## How to customize → `{{STORE_AND_CATALOG}}`: what you sell and how big the catalog is, for example "outdoor gear retailer, about 8,000 SKUs across tents, packs, and footwear" → `{{SITE_STRUCTURE}}`: your URL taxonomy, for example "/category/subcategory/product, plus /blog and /guides" → `{{CONVERSION_GOAL}}`: the goal and current target, for example "purchases at a 350% ROAS floor" → `{{EXISTING_CAMPAIGNS}}`: what already runs and performs, so the plan complements rather than cannibalizes → `{{AUDIENCES}}`: remarketing lists, customer lists, or site visitor segments you can layer for a safe launch → `{{BUDGET}}`: the monthly amount you are willing to expose to this test → `{{EXCLUSIONS}}`: page or category types to keep out, for example "low price accessories, clearance, blog, policy pages" ## What good output looks like → The fit verdict is decisive: it either commits to DSA against all three criteria or recommends a keyword led structure instead, with reasoning, rather than hedging → The page feed table names Custom Labels you can actually optimize on later, and it visibly excludes every protected page type you listed → The launch plan is layered on a real audience, starts narrow, starts bidding open, and negates the exact match queries your existing campaigns already win, so the first weeks are instrumented rather than a budget gamble → The weekly runbook covers all three Search Terms views and specifies negative dynamic ad targets by URL or rule, so poorly converting pages Google chose get shut off fast → Description copy carries a specific differentiator, fits the stated character limit, and reads as a finished ad, with page title improvements called out separately since Google writes the headlines → The answer ends with a fact check list of platform specifics (report names, feed format, character limits, settings names) flagged for verification in the current Google Ads interface ## Related prompts → [Audit Google Ads for Wasted Spend](../paid-search/audit-google-ads-wasted-spend.md): run this after launch to catch the queries and pages DSA started converting on that you never intended → [Structure a Google Ads Account](../paid-search/structure-a-google-ads-account.md): decide where the DSA campaign sits relative to your keyword led campaigns before you build it → [Choose Shopping vs Performance Max](../paid-search/choose-shopping-vs-performance-max.md): weigh DSA against the other automation surfaces when deciding how to harvest long tail demand at scale