Adspirer Ads

Manage advertising campaigns across Google Ads, Meta Ads, LinkedIn Ads, and TikTok Ads using the Adspirer MCP server (400+ tools across Google Search/PMax/Display/Demand Gen/YouTube, Meta image/video/carousel/lead-gen, LinkedIn sponsored content/carousel/lead-gen with campaign groups, and TikTok in-feed/Spark/Carousel/App Promotion).

When to Use This Skill

Activate when the user:

• Asks about ad campaign performance ("How are my Google Ads doing?")
• Wants to research keywords ("Find keywords for my plumbing business")
• Needs to create campaigns ("Launch a Google Search campaign for...")
• Wants budget optimization ("Where am I wasting ad spend?")
• Mentions advertising platforms (Google Ads, Meta, LinkedIn, TikTok)
• Asks about ad accounts or connections ("Which ad platforms are connected?")

Required Workflow

Follow these steps in order. Do not skip steps.

Step 1: Check Connected Platforms

Always start here before any ad operation:

• Call get_connections_status
• Shows connected platforms, primary/secondary accounts, account IDs
• If the target platform is not connected, direct the user to https://adspirer.ai/connections

Step 2: Identify the Task

Step 3: Execute Tools

Follow the workflow patterns below. Always read first (performance, status), then act (create, optimize).

Step 4: Summarize and Recommend

Present results in tables with key metrics. Highlight top and underperforming items. Propose actionable next steps.

Performance Analysis

• Google Ads: get_campaign_performance — params: lookback_days (7/30/60/90, default 30), optional customer_id
• Meta Ads: get_meta_campaign_performance — params: lookback_days, optional ad_account_id
• LinkedIn Ads: get_linkedin_campaign_performance — params: lookback_days
• TikTok Ads: get_tiktok_campaign_performance — params: lookback_days

Present: impressions, clicks, CTR, spend, conversions, cost/conversion, ROAS. Default to 30-day lookback.

Cross-Platform Performance Dashboard

When the user asks for overall performance, a weekly review, or cross-platform comparison:

1. Call get_connections_status to identify active platforms
2. For each connected platform, pull performance:
   • Google: get_campaign_performance
   • LinkedIn: get_linkedin_campaign_performance
   • Meta: get_meta_campaign_performance
   • TikTok: get_tiktok_campaign_performance
3. For each platform, pull waste analysis:
   • Google: analyze_wasted_spend
   • LinkedIn: analyze_linkedin_wasted_spend
   • Meta: analyze_meta_wasted_spend
   • TikTok: analyze_tiktok_wasted_spend
4. Present a unified scorecard:

5. Highlight:
   • Best performing platform and campaign
   • Worst performing platform and campaign
   • Total wasted spend and top waste sources
   • Budget pacing (on track, under, over)
6. Recommend top 3 actions across all platforms

Campaign Research (run before creating ANY new campaign)

Before creating a campaign on any platform, research the brand's market position and competitive landscape. This combines web research with ad platform data (Adspirer MCP) to inform every campaign decision — targeting, messaging, differentiation, and bidding.

Step 0: Load Strategy Directives

Read STRATEGY.md — ## Active Directives and skim ## Decision Log. If directives exist, note them as context that will inform (not skip) the research steps that follow. Directives shape which competitors to focus on and what positioning angles to explore.

Do NOT skip Campaign Research just because directives exist. Directives may be stale, incomplete, or based on exploratory conversations. Fresh research validates and enriches them. However, avoid fully redundant research — if a comprehensive analysis was done recently (check Decision Log dates), tell the user: "Strategy directives from [date] are available. I'll use them as a starting point and validate with fresh data. Want me to do a full re-analysis instead?"

Step 1: Understand the brand's own website

Crawl the brand's website. Extract:

• What they sell (products, services, pricing tiers)
• Key value propositions and differentiators
• Target audience language (how they describe their customers)
• Pricing (plans, tiers, free trial availability)
• Trust signals (customer logos, testimonials, case studies, awards)
• CTAs used on the site (what actions they push visitors toward)

Step 2: Research the competitive landscape

Search for:

• "[brand's product category] competitors" — find who they compete with
• "[competitor name] vs [brand name]" — find comparison content
• "[competitor name] pricing" — understand competitor price points
• "best [product category] [current year]" — find review/comparison sites

Then crawl the top 3-5 competitor websites. Extract:

• Their positioning and messaging (how they describe themselves)
• Their pricing (cheaper? more expensive? different model?)
• Their unique claims (what do they say they do better?)
• Their target audience (who are they speaking to?)

Step 3: Identify differentiation

Combine brand website + competitor research to answer:

• What does this brand do that competitors don't? (unique features, approach, pricing)
• What language resonates in this market? (common pain points, desired outcomes)
• Where are the gaps? (underserved audiences, unaddressed pain points)
• What should ad copy emphasize to stand out?

Step 4: Pull existing ad intelligence from Adspirer

• get_campaign_performance — what's already running and how it performs
• analyze_search_terms — what real users search for (Google Ads)
• get_campaign_structure — current ad copy and targeting
• get_benchmark_context — industry benchmarks for this vertical

Step 5: Create a research brief

Present findings to the user before proceeding with campaign creation:

• Market overview (key competitors, price ranges, positioning)
• Recommended differentiation angles (what to emphasize in ads)
• Suggested audiences based on competitive gaps
• Messaging direction (informed by competitor weaknesses and brand strengths)

Get user input on the direction before proceeding to keyword research and campaign creation.

Keyword Research (Google Ads)

Always run before creating Search campaigns. Never use generic SEO keywords.

• Tool: research_keywords
• Params: business_description or seed_keywords, optional website_url, target_location
• Group results by intent (high/medium/low), show search volume, CPC ranges, competition
• Use insights from Campaign Research to inform seed keywords — include competitor brand terms, differentiation keywords, and pain-point language discovered during research
• Strategy directive filter (MANDATORY): After research_keywords returns results, cross-reference against STRATEGY.md > Google Ads directives:
  • Deprioritize keywords matching AVOID directives. Note these to the user.
  • Highlight keywords matching PREFER directives.
  • Apply CONSTRAINT directives (match type rules, budget caps).
  • Flag conflicts: "Keyword '[term]' returned by research but conflicts with active directive: '[text]'. Deprioritizing it."

Bidding Strategy

Before creating ANY Google Ads campaign, discuss bidding strategy with the user.

1. Pull past performance: get_campaign_performance (lookback_days: 90)
2. Review existing strategies: get_campaign_structure to see what bidding strategies current campaigns use
3. Recommend a strategy based on data:

4. Present recommendation with reasoning to the user
5. Get explicit approval before setting the strategy
6. To change strategy on existing campaigns: update_bid_strategy

Important: Never silently pick a bidding strategy. Always explain the trade-offs and let the user decide.

Campaign Creation

For ALL new campaigns: Run Campaign Research first (see section above) unless the user has already provided competitive context or this is a follow-up campaign for an existing brand workspace with research already done.

Google Ads Search (exact order):

1. Campaign Research — crawl brand + competitor websites, present research brief 1.5. Apply strategy directives — load STRATEGY.md. Use as context for keyword selection, bidding, targeting, and ad copy throughout this creation flow.
2. research_keywords — mandatory, informed by competitive research
3. Discuss bidding strategy with user (see Bidding Strategy section above)
4. discover_existing_assets — check for existing ad assets
5. validate_and_prepare_assets — validate before creation (use differentiation angles from research in ad copy)
6. create_search_campaign — create the campaign (PAUSED status)
7. Add ad extensions (see Ad Extensions section below):
   • Crawl user's website to find real page URLs
   • add_sitelinks — add 10+ validated sitelinks
   • add_callout_extensions — add 4+ callouts (use value props from research)
   • add_structured_snippets — add relevant structured snippets
8. list_campaign_extensions — verify all extensions were added
9. Run post-create verification on this campaign using get_campaign_structure:
   • confirm ad groups exist
   • confirm keywords exist with expected match-type profile
   • confirm at least one RSA exists
10. If any required asset is missing, run one targeted remediation pass for the missing asset class only, then re-verify.
11. Do not report success until this campaign passes Launch Definition of Done.

Google Ads Performance Max:

1. Campaign Research — crawl brand + competitor websites, present research brief 1.5. Apply strategy directives — load STRATEGY.md. Use as context for bidding, targeting, and creative direction throughout this creation flow.
2. Discuss bidding strategy with user (see Bidding Strategy section above)
3. discover_existing_assets — check existing assets
4. validate_and_prepare_assets — validate creative assets (see PMax Asset Limits below)
5. validate_video — validate YouTube videos if user provides them (see PMax Asset Limits)
6. create_pmax_campaign — create the campaign
7. Add ad extensions (same as Search — sitelinks, callouts, snippets)
8. list_campaign_extensions — verify all extensions were added
9. Add search themes (recommended) — see PMax Search Themes & Audience Signals section below:
   • Ask user for search themes or derive from keyword research + brand context
   • add_pmax_search_themes — add up to 50 themes per asset group
   • get_pmax_search_themes — verify themes were added
10. Add audience signals (recommended) — see PMax Search Themes & Audience Signals section below:
    • search_audiences — find relevant in-market, affinity, and custom audiences
    • Present audience recommendations to user for approval
    • add_pmax_audience_signal — add audience signal combining selected segments

Google Ads Demand Gen (exact order):

Demand Gen campaigns run across YouTube (In-Feed, In-Stream, Shorts), Gmail, Discover, and Display. They support two ad formats: multi_asset (image ads) and video_responsive (video ads). Demand Gen is ideal for awareness and consideration campaigns with visual, entertainment-focused creatives.

1. Campaign Research — crawl brand + competitor websites, present research brief
2. Discuss bidding strategy with user:
   • Maximize Clicks (default, no conversion tracking needed)
   • Target CPA (requires conversion tracking)
   • Maximize Conversions (requires conversion tracking)
   • Target ROAS (requires conversion value tracking)
3. discover_existing_assets — check existing images, logos, videos
4. For multi_asset format: need landscape (1.91:1) OR square (1:1) images + logo
5. For video_responsive format: need YouTube video IDs (validate with validate_video) + logo
6. search_audiences — find relevant in-market, affinity audiences for audience_segments parameter
7. create_demandgen_campaign — create the campaign with:
   • target_locations: supports countries, states, cities globally (e.g., ["India"], ["Bangalore, India"])
   • audience_segments: {"in_market_audience_ids": [80463], "affinity_audience_ids": [92913]}
   • channels: optional channel controls (e.g., {"display": false} for YouTube + Gmail + Discover only)
   • call_to_action: e.g., "LEARN_MORE", "SIGN_UP", "SHOP_NOW"
   • existing_images: reuse from discover_existing_assets (preferred)
8. Add ad extensions (sitelinks, callouts, snippets)
9. list_campaign_extensions — verify extensions added

Key Demand Gen details:

• Budget: minimum ~$20/day recommended
• All campaigns created PAUSED
• Location and audience targeting are set at the ad group level (not campaign level)
• Audience targeting uses grouped Audience resources with in-market, affinity, and remarketing segments
• Channel controls: all channels ON by default. Disable specific channels via channels parameter

Google Ads YouTube Video (exact order):

YouTube campaigns are Demand Gen campaigns with YouTube-only channel controls (Gmail, Discover, Display disabled). They require at least one YouTube video.

1. Get user's YouTube video URL or ID
2. validate_video — validate the video (must be public or unlisted)
3. discover_existing_assets — find existing logos
4. search_audiences — find relevant audiences
5. create_youtube_campaign — create with:
   • youtube_video_id: primary video (11 chars)
   • target_locations: city/state/country level
   • audience_segments: in-market + affinity targeting
   • logo_asset_id OR asset_bundle_id for logo
   • additional_video_ids: up to 4 more videos
6. Add ad extensions (sitelinks, callouts, snippets)

YouTube placements enabled: In-Feed, In-Stream, Shorts YouTube placements disabled: Gmail, Discover, Display

Google Ads Display (Standard + Smart Display):

1. Campaign Research — crawl brand + competitor sites; identify visual positioning angles 1.5. Apply strategy directives — load STRATEGY.md. Use for targeting and creative direction.
2. Discuss bidding strategy with user
3. select_google_campaign_type — confirm "display" branch (Standard vs Smart Display)
4. resolve_google_locations + list_google_languages — resolve geo + language inputs (rejects ambiguous strings)
5. discover_existing_assets + validate_and_prepare_assets
6. create_display_campaign — created PAUSED by default; single-flag toggle to launch live
7. add_display_ad_group — add ad groups (with optional schedule + frequency caps)
8. Targeting via per-surface ADD tools — audiences, topics, placements, keywords, demographics. Use remove_display_criteria for removals.
9. add_display_ad — Responsive Display Ads (RDA). Later edits via update_ad_creative / update_ad_headlines / update_ad_descriptions (router auto-dispatches RSA vs RDA)

Meta Ads:

1. Campaign Research — crawl brand + competitor websites, understand audience positioning 1.5. Apply strategy directives — load STRATEGY.md > Meta Ads and Cross-Platform sections. Use as context for audience targeting and creative direction.
2. get_connections_status — verify Meta account connected
3. select_meta_campaign_type — pick objective (TRAFFIC, OUTCOME_LEADS, OUTCOME_ENGAGEMENT, CONVERSIONS, app)
4. search_meta_targeting / browse_meta_targeting — find interest, behavior, demographic, geo (city-level supported), language; list_meta_custom_audiences for retargeting / lookalikes
5. discover_meta_assets + validate_and_prepare_meta_assets — validate creative
6. Create campaign — pick format:
   • create_meta_image_campaign — single-image; supports placement-specific creatives (Feed / Stories / Reels) and emoji headlines
   • create_meta_video_campaign — video creative (DCO supported)
   • create_meta_carousel_campaign — carousel cards
   • For OUTCOME_LEADS: pass lead_form_id (auto-fetched via list_meta_lead_forms if omitted)
   • App campaigns: app_link_spec is wired end-to-end
7. Optional: add_meta_ad_set for additional ad sets — supports lifetime budgets, end_time, granular placements (publisher_platforms, facebook_positions, instagram_positions), multi_advertiser, custom_conversion_id, Advantage+ Audience and Advantage+ Creative opt-outs (degrees_of_freedom_spec)

LinkedIn Ads:

1. Campaign Research — crawl brand + competitor websites, understand B2B positioning 1.5. Apply strategy directives — load STRATEGY.md > LinkedIn Ads and Cross-Platform sections. Use as context for targeting, messaging, and budget allocation.
2. get_linkedin_organizations — get linked company pages
3. search_linkedin_targeting (typeahead) or research_business_for_linkedin_targeting — 14 facets (job functions, seniority, industries, company size, skills, member groups)
4. discover_linkedin_assets — check existing creatives (paginated for 22k+ creatives, deduped)
5. validate_and_prepare_linkedin_assets — validate single-image, video, or carousel assets
6. Create campaign — pick format:
   • create_linkedin_image_campaign — single-image sponsored content
   • create_linkedin_video_campaign — video sponsored content
   • create_linkedin_carousel_campaign — carousel; build cards via add_linkedin_carousel_creative
   • For lead gen: pass lead_gen_form_id
7. Optional: campaign groups via campaign_group_id / campaign_group_name or add_linkedin_campaign_to_group. list_linkedin_campaigns filters by group.

TikTok Ads:

1. Campaign Research — crawl brand website, research competitor video ad strategies 1.5. Apply strategy directives — load STRATEGY.md > TikTok Ads and Cross-Platform sections. Use as context for creative direction and audience targeting.
2. discover_tiktok_assets — check existing video / image assets
3. validate_and_prepare_tiktok_assets — validate creative assets
4. search_tiktok_targeting — find interests, behaviors, geo (city-level)
5. Create campaign — pick objective:
   • create_tiktok_campaign — flexible objective; pass tiktok_item_id or card_id for Spark Ads (boost organic)
   • create_tiktok_video_campaign — in-feed video
   • APP_PROMOTION objective for app installs
   • For carousel: create_tiktok_carousel_card then add_tiktok_ad
6. Optional: CBO via budget_optimize_on (default true). add_tiktok_ad_group for additional ad groups; update_tiktok_* to adjust live campaigns
7. CTAs are server-validated (defaults to LEARN_MORE if landing_page_url provided without call_to_action). optimization_event accepts the full event taxonomy: FORM, ON_WEB_CART, ON_WEB_DETAIL, ON_WEB_REGISTER, COMPLETE_PAYMENT, CONVERSION_LEADS, PAGE_VISIT, etc.

Ad Extensions (Google Ads)

Ad extensions improve Quality Score, increase ad real estate, and boost CTR. Always add extensions after creating a Google Ads campaign.

Before Adding Extensions

Call list_campaign_extensions to check what already exists on the campaign. Never duplicate existing extensions.

Sitelinks (add_sitelinks)

Sitelinks are the most impactful extension. Target 10+ sitelinks (more is better — Google rotates the best performers).

Workflow:

1. Crawl the user's website homepage — extract all navigation links and key pages
2. Build a candidate list of pages to include:
   • Homepage, Pricing/Plans, About Us, Contact, Key product/service pages, Blog, Case Studies/Testimonials, FAQ/Help, Free Trial/Demo, Careers
3. Validate each URL — confirm it loads (no 404s, no redirects to error pages)
4. For each valid sitelink, prepare:
   • Link text: max 25 characters, descriptive (e.g., "View Pricing Plans")
   • Description line 1: max 35 characters (e.g., "Plans starting at $29/month")
   • Description line 2: max 35 characters (e.g., "Free 14-day trial included")
   • Final URL: the validated page URL
5. If fewer than 8 valid pages found → ask the user for additional URLs or pages to include
6. Present the full sitelink list to the user for review before adding
7. Call add_sitelinks with the approved list

Callout Extensions (add_callout_extensions)

Callouts highlight value propositions. Target 8+ callouts (minimum 4).

Workflow:

1. Extract value propositions from:
   • Website content (crawled)
   • Brand docs (AGENTS.md if it exists)
   • Existing ad copy in the account
2. Each callout: max 25 characters
3. Examples: "Free Shipping", "24/7 Support", "No Setup Fee", "Cancel Anytime", "Money-Back Guarantee", "Same-Day Delivery", "Award-Winning", "Trusted by 10K+"
4. Present to user for approval, then call add_callout_extensions

Structured Snippets (add_structured_snippets)

Snippets show predefined categories of offerings. Pick headers relevant to the business.

Available headers: Brands, Courses, Destinations, Featured Hotels, Insurance Coverage, Models, Neighborhoods, Service Catalog, Shows, Styles, Types

Workflow:

1. Review the user's website and business type to pick relevant headers
2. Extract 3-10 values per header from website content
3. Example: SaaS company → Header "Types" with values "Analytics, Reporting, Dashboards, API Access"
4. Example: E-commerce → Header "Brands" with values "Nike, Adidas, Puma, New Balance"
5. Present to user, then call add_structured_snippets

Price Extensions

If the user's website has a pricing page:

1. Crawl the pricing page
2. Extract plan names, prices, and descriptions
3. Present to user for confirmation before adding

Call Extensions

If the business has a phone number:

1. Ask the user for their business phone number
2. Discuss call tracking preferences
3. Set call hours if business has limited availability

Extension Verification

After adding all extensions, always call list_campaign_extensions to verify:

• All sitelinks were added and have correct URLs
• Callouts are present
• Structured snippets are showing
• Report back to the user what was added

If list_campaign_extensions fails:

1. Retry once with a corrected payload or narrower scope.
2. Cross-check extension state with get_campaign_structure.
3. If extension state is still unverifiable, report PARTIAL_SUCCESS and explicitly list what could not be confirmed.

Launch Definition of Done (Required before reporting success)

For each created campaign, all checks below must pass:

1. Campaign exists and status is PAUSED (unless user explicitly approved activation).
2. Expected ad group count exists.
3. Expected keywords exist, with the planned match-type profile (EXACT/PHRASE/BROAD as specified).
4. At least one RSA exists with expected headline/description counts.
5. Required extensions exist (sitelinks, callouts, structured snippets for Google Ads campaigns).
6. Requested-vs-actual bidding strategy matches, or any drift is explicitly called out.

Status protocol (mandatory)

• SUCCESS: all Launch Definition of Done checks pass for all targeted campaigns.
• PARTIAL_SUCCESS: campaign shell exists, but one or more required assets are missing/unverifiable after one targeted remediation pass.
• FAILED: campaign creation itself failed.

Never report SUCCESS when any verification check failed or could not be confirmed.

Per-campaign action ledger (mandatory)

For every campaign you create or edit, log and return:

• campaign name + campaign_id
• ad_group_id values touched
• keyword add/update counts
• RSA counts (headlines/descriptions)
• extension counts (sitelinks/callouts/snippets)
• verification result per campaign (PASS/FAIL)

Ad Quality Guardrails (Google Ads)

Keyword-to-headline coverage

For each ad group, include unique high-intent target keywords in RSA headlines. At minimum:

• cover top 3-5 keyword themes from the ad group's keyword list
• avoid generic headline sets that omit core search intent terms
• when competitive terms are targeted, keep competitor names in keywords only if required by strategy and avoid naming competitors in ad copy unless user explicitly approves

Asset length validation checklist (before submission)

• RSA headline: <= 30 characters
• RSA description: <= 90 characters
• Path fields: <= 15 characters each
• Sitelink text: <= 25 characters
• Sitelink description lines: <= 35 characters each
• Callout text: <= 25 characters
• Structured snippet values: follow platform limits and keep concise

Validate lengths before create_* or add_* calls. If limits are exceeded, rewrite and re-validate before submitting.

PMax Asset Limits (CRITICAL — enforce before every PMax campaign)

Google Ads Performance Max enforces strict asset limits per asset group. Exceeding these limits causes campaign creation to fail with ENABLED_IMAGE_ASSET_LINKS_PER_ASSET_GROUP or similar errors. The agent MUST enforce these limits before calling validate_and_prepare_assets or create_pmax_campaign.

Image Asset Limits

CRITICAL: The 20-image limit is TOTAL across landscape + square + portrait, NOT 20 per ratio. If a user provides 10 landscape + 15 square + 13 portrait = 38 images, you must reduce to 20 total.

When user provides more than 20 total marketing images:

1. Inform the user: "Google Ads allows max 20 marketing images total per PMax asset group across all ratios. You provided [N]. I'll select the best 20."
2. Distribute proportionally across ratios while respecting minimums (at least 1 landscape, at least 1 square)
3. Prefer keeping variety across ratios over loading up one ratio
4. validate_and_prepare_assets handles this automatically, but you should inform the user proactively

When user provides fewer than minimums:

• Missing landscape images: Ask the user to provide at least 1 landscape image (1.91:1 ratio, 1200x628px recommended)
• Missing square images: Ask the user to provide at least 1 square image (1:1 ratio, 1200x1200px recommended)
• Missing square logo: Ask the user to provide at least 1 square logo (1:1 ratio, 128x128px minimum)
• Do NOT proceed with campaign creation until minimums are met

Text Asset Limits

When user provides more than the maximum:

• Truncate to the limit, keeping the first N items
• Inform the user: "PMax allows max [N] [asset type]. Using the first [N]."

When user provides fewer than minimums:

• Ask for the missing assets. Example: "PMax requires at least 3 headlines. You provided 1. Please provide 2 more headlines (max 30 characters each)."
• Do NOT proceed until minimums are met

Video Asset Limits

Video validation workflow:

1. Call validate_video with platform="pmax" for each video
2. This validates the YouTube video ID format only (11 chars, alphanumeric)
3. Google Ads API will verify the video exists, is public/unlisted, and embeddable during campaign creation
4. If user provides more than 5 videos, use only the first 5 and inform them
5. Videos are optional — campaigns can be created without them (Google will auto-generate video ads from images)

Asset Limit Enforcement Checklist (before calling create_pmax_campaign)

Before creating any PMax campaign, verify:

• At least 1 landscape image provided
• At least 1 square image provided
• Total marketing images (landscape + square + portrait) <= 20
• At least 1 square logo provided
• At least 3 headlines (max 30 chars each)
• At least 1 long headline (max 90 chars)
• At least 2 descriptions (max 90 chars each)
• Business name provided (max 25 chars)
• Videos (if any) validated via validate_video
• Videos count <= 5

If any check fails, ask the user for the missing/corrected assets before proceeding.

PMax Search Themes & Audience Signals

Search themes and audience signals are PMax-only features — they do not apply to Search, Display, or other campaign types. Both are configured at the asset group level.

Search Themes

Search themes are short phrases (max 50 per asset group) that tell Google which search queries your PMax campaign should target. They supplement Google's automated targeting with explicit intent signals.

When to add search themes:

• During PMax campaign creation (step 9 above)
• When user wants to refine PMax targeting ("add search themes to my PMax campaign")
• When search term analysis reveals high-performing queries not yet covered
• When expanding into new market segments

Workflow:

1. get_pmax_search_themes — check existing themes on the campaign (requires campaign_id)
2. Derive theme candidates from:
   • Keyword research results (research_keywords)
   • Search term reports (analyze_search_terms)
   • Brand context and product categories
   • Competitive research findings
3. Present candidate themes to user for approval
4. add_pmax_search_themes — add approved themes (params: campaign_id, search_themes list)
5. get_pmax_search_themes — verify themes were added

Limits and rules:

• Max 50 search themes per asset group (Google's limit)
• Duplicates are filtered automatically (case-insensitive)
• Empty/whitespace themes are filtered automatically
• Search themes only support add and remove — no update (remove + re-add instead)
• To remove: remove_pmax_search_themes with the theme resource names from get_pmax_search_themes

Audience Signals

Audience signals tell Google which audience segments are most likely to convert. They are signals, not hard targeting — Google uses them as starting points and expands from there. Only one audience signal is allowed per asset group (containing multiple segments).

When to add audience signals:

• During PMax campaign creation (step 10 above)
• When user wants to add targeting to a PMax campaign
• When audience analysis reveals high-value segments
• When launching PMax for a specific product/audience niche

Supported segment types:

Workflow:

1. get_pmax_audience_signals — check existing signals on the campaign (requires campaign_id)
2. Discover available audiences:
   • search_audiences — search by keyword across all audience types (in-market, affinity, custom)
   • Use brand context and competitive research to inform search terms
3. Present audience recommendations organized by type with rationale
4. Get user approval for segment selection
5. add_pmax_audience_signal — add signal with selected segment IDs
6. get_pmax_audience_signals — verify signal was added

Limits and rules:

• Only one audience signal per asset group — it must combine all desired segments
• To change: remove_pmax_audience_signal (with resource name), then add_pmax_audience_signal with updated segments
• Audience signals only support add and remove — no update
• Use search_audiences to find audience IDs — do not guess or hardcode IDs

Conversion Tracking Limitation

Adspirer MCP currently does not configure Google Ads conversion action settings (primary vs secondary) directly.

When campaign goals depend on conversion action priority:

1. create campaigns in PAUSED status
2. tell the user exact Google Ads UI path to configure conversion actions manually
3. report campaign creation as complete only after clarifying this manual step

Budget Optimization (Google Ads)

• optimize_budget_allocation — suggest budget reallocations
• analyze_wasted_spend — find underperforming keywords and ad groups
• analyze_search_terms — review search terms for negative keyword opportunities

Creative Fatigue Detection & Refresh

When reviewing creative performance or user asks about ad fatigue:

1. Meta: Call detect_meta_creative_fatigue for fatigue scores
2. LinkedIn: Call analyze_linkedin_creative_performance for per-creative metrics
3. Google: Call get_campaign_structure to see ad-level performance
4. Identify ads with:
   • High frequency + declining CTR (fatigued)
   • More than 30 days running with no refresh
   • Below-average CTR for their campaign
5. For fatigued ads:
   • Call suggest_ad_content for new headline/description ideas
   • Call generate_linkedin_ad_creatives for LinkedIn variations
   • Present new creative options (filtered through brand voice if AGENTS.md exists)
   • On approval: update via update_ad_headlines, update_ad_descriptions, add_linkedin_creative, etc.

A/B Testing Workflow

When creating new ad variations for testing:

1. Read current top-performing ad copy (from campaign structure)
2. Generate 3-5 variations testing ONE variable:
   • Headline variation (keep description same)
   • Description variation (keep headline same)
   • CTA variation
   • Audience variation (same ad, different targeting)
3. Present test plan with hypothesis: "Testing: 'Headline A' vs 'Headline B' Hypothesis: [why we think one may outperform] Duration: 2 weeks, split budget 50/50 Success metric: CTR and conversion rate"
4. On approval, create test variants via platform-specific tools
5. Log test for follow-up analysis

Audience Analysis & Optimization

When analyzing or optimizing audiences:

1. Pull audience data:
   • Meta: get_meta_audience_insights + analyze_meta_audiences
   • LinkedIn: get_linkedin_audience_insights
   • Google: Review campaign targeting from get_campaign_structure
2. Identify:
   • Which audience segments convert best (by age, gender, job title, interest)
   • Audience overlap/saturation
   • Underperforming segments (high spend, low conversion)
3. For LinkedIn B2B specifically:
   • Call research_business_for_linkedin_targeting with brand's website
   • Compare AI-recommended targeting vs current targeting
   • Identify gaps (seniority levels, industries, company sizes not covered)
4. For Meta:
   • Call optimize_meta_placements for placement-level performance
   • Identify which placements to scale/reduce
5. Present findings with recommendations:
   • Segments to expand (high ROAS, low spend)
   • Segments to cut (low ROAS, high spend)
   • New segments to test

Monitoring & Alerts

When user wants alerts or ongoing monitoring:

1. Understand what they want to monitor:
   • Budget pacing (approaching monthly limit)
   • Performance drops (ROAS below target, CPA above target)
   • Opportunity alerts (keyword with sudden volume increase)
2. Call create_monitor with appropriate:
   • Metric (ROAS, CTR, CPC, CPA, spend, conversions)
   • Threshold and direction (below 3.0, above $150)
   • Delivery (email, Slack, SMS)
3. Call list_monitors to confirm setup
4. To modify: manage_scheduled_task with monitor ID

Reporting

When user wants performance reports:

1. Ask: frequency (one-time, weekly, monthly) and format preference
2. For one-time: Call generate_report_now
3. For recurring: Call schedule_brief with:
   • Frequency (daily, weekly)
   • Delivery method (email, Slack, webhook)
   • Content scope (all platforms or specific)
4. Call list_scheduled_tasks to confirm

Competitive Intelligence

When analyzing competitors or adjusting competitive strategy:

Step 1: Identify competitors

• Read AGENTS.md for known competitors (check Competitors section)
• Sear

…(truncated)