Optimization

Add audiences in observation mode to Search campaigns and learn who converts best

Adds your own lists and Google segments to Search campaigns in observation mode without narrowing reach. Checks the targeting mode trap first, reads the result back after adding and leaves a reading plan for 30 days later.

When to use it

When you do not know which audiences your Search campaigns work best for and want to start collecting that data without cutting reach.

What you get

Targeting mode warnings, audiences added in observation mode per campaign and confirmed by reading them back, and a ready reading plan that shows in 30 days which audiences stand out.

Does it change anything in my account?

Only with your approval. The assistant first shows a preview of every change and waits for your go-ahead.

Works with

Google Ads

Fill in before sending

[account name][business and customer description]

I want to add audience segments to the Search campaigns in my Google Ads account [account name] without narrowing their reach. The goal is to see which audiences convert better and decide later based on that. This is how I would describe our business and customer: [business and customer description]. 1. Current state. Get the enabled Search campaigns, their bidding strategies and last 30 days' results with account_snapshot. With google_ads_query, pull the audiences already attached from campaign_criterion and ad_group_criterion (type IN USER_LIST, USER_INTEREST, CUSTOM_AUDIENCE, COMBINED_AUDIENCE, with the negative field). 2. Read the targeting mode. With google_ads_query, pull targeting_setting.target_restrictions from the campaign and ad_group resources and look at the AUDIENCE dimension. If bid_only=false, that level is in "Targeting" mode and every audience added there narrows reach. Show these campaigns and ad groups first and ask me before adding anything to them. 3. My own lists. With list_audiences and definition=true, get the lists, their types and their Search network sizes. Google needs at least 1,000 active members for a list to be used on Search; leave lists below that out of the proposal and show them separately. 4. Proposal. Propose a set of at most 10 audiences per campaign: my own lists such as site visitors and converters, in-market segments that fit my business and, where useful, one detailed demographic segment. Write Google segments with their exact name as shown in the Google Ads interface. Do not propose life event segments for Search campaigns, because Google does not accept them on Search. 5. Apply. After my approval, add each audience with manage_audience_targeting using action=attach and mode=OBSERVATION. Never leave mode empty and do not set bid_modifier. Preview all additions per campaign with confirm=false first, then apply them, then read campaign_criterion and targeting_setting again with google_ads_query to confirm the audiences were added and the targeting mode did not change. Also write a reading plan for me to use in 30 days: audience-level impressions, clicks, cost and conversions from campaign_audience_view and ad_group_audience_view with google_ads_query. Decision rule: do not judge an audience with fewer than 100 clicks in 30 days; flag audiences whose conversion rate is 30% above or below the campaign average. Note that audience bid adjustments are not applied on Smart Bidding campaigns, so findings there need to turn into ad copy, separate campaign or exclusion decisions instead of bid changes. Format the output like this: targeting mode warnings, a table of proposed audiences per campaign (campaign, audience, type, Search size, reason) and the 30-day reading plan.

1. Current state. Get the enabled Search campaigns, their bidding strategies and last 30 days' results with account_snapshot. With google_ads_query, pull the audiences already attached from campaign_criterion and ad_group_criterion (type IN USER_LIST, USER_INTEREST, CUSTOM_AUDIENCE, COMBINED_AUDIENCE, with the negative field).
2. Read the targeting mode. With google_ads_query, pull targeting_setting.target_restrictions from the campaign and ad_group resources and look at the AUDIENCE dimension. If bid_only=false, that level is in "Targeting" mode and every audience added there narrows reach. Show these campaigns and ad groups first and ask me before adding anything to them.
3. My own lists. With list_audiences and definition=true, get the lists, their types and their Search network sizes. Google needs at least 1,000 active members for a list to be used on Search; leave lists below that out of the proposal and show them separately.
4. Proposal. Propose a set of at most 10 audiences per campaign: my own lists such as site visitors and converters, in-market segments that fit my business and, where useful, one detailed demographic segment. Write Google segments with their exact name as shown in the Google Ads interface. Do not propose life event segments for Search campaigns, because Google does not accept them on Search.
5. Apply. After my approval, add each audience with manage_audience_targeting using action=attach and mode=OBSERVATION. Never leave mode empty and do not set bid_modifier. Preview all additions per campaign with confirm=false first, then apply them, then read campaign_criterion and targeting_setting again with google_ads_query to confirm the audiences were added and the targeting mode did not change.

Also write a reading plan for me to use in 30 days: audience-level impressions, clicks, cost and conversions from campaign_audience_view and ad_group_audience_view with google_ads_query. Decision rule: do not judge an audience with fewer than 100 clicks in 30 days; flag audiences whose conversion rate is 30% above or below the campaign average. Note that audience bid adjustments are not applied on Smart Bidding campaigns, so findings there need to turn into ad copy, separate campaign or exclusion decisions instead of bid changes.

Format the output like this: targeting mode warnings, a table of proposed audiences per campaign (campaign, audience, type, Search size, reason) and the 30-day reading plan.

Open in Claude Open in ChatGPT Discuss in the forum

Example result

An example of the output this prompt produces. Its structure comes from the tools' real output; numbers and names are fictional.

Example output (based on a fictional account, not real customer data)

Targeting mode warning: the AUDIENCE dimension on "Search South Region" has bid_only=false, so it is in Targeting mode. Adding audiences there would show ads only to those audiences. I left this campaign out until you approve.

Campaign	Audience	Type	Search size	Reason
Search Services	All visitors (30 days)	Own list	4,200	To see how returning visitors convert
Search Services	Converters (180 days)	Own list	1,350	To see how often existing customers click
Search Services	Home Improvement Services	In-market		Directly related to the service
Search Services	Homeowners	Detailed demographics		To separate tenants from owners

Left out of the proposal: "Blog readers", Search size 640, at least 1,000 is needed.

30-day reading plan: pull clicks, cost and conversions per audience from campaign_audience_view and ad_group_audience_view. Audiences under 100 clicks will not be judged. The campaign uses Target CPA, so standout audiences will be acted on with separate ad copy or a separate campaign, not bid adjustments.

More prompts

OptimizationChanges after your approval

Why are my ads not showing? First page bid, rank and budget diagnosis

Separates, keyword by keyword, whether low spend or few impressions come from bids, quality or budget. Applies bid and budget changes only after your approval.

When to use itWhen your Google Ads campaigns spend far below budget or barely get impressions and you need to know whether bids, quality or budget is the cause.

In my Google Ads account [account name], [campaign names or "all Search campaigns"] spend far below budget or get very few impressions. Use the last 30 days to find out why.

1. Campaign level. Use impression_share_analysis to show each campaign's impression share, the share lost to budget and to rank, and how often the ads appear at the top of the page. If a campaign cannot fill its budget yet still shows share lost to budget, explain the likely reason.
2. Keyword level. With keyword_quality_report and google_ads_query on keyword_view, pull each active keyword's bid, first page and top of page estimates, Quality Score and its three components. Sort the keywords into four groups: bid below the first page estimate, rarely shown because quality is low, volume too low, no problem.
3. Recommendation. For keywords whose bid is too low and that fit my business, propose a new bid. It must not be below the first page estimate and must not exceed [max cost per click]. For low-quality keywords, instead of raising the bid, say which component needs fixing: ad relevance, expected click-through rate or landing page. For campaigns that lose share to budget and bring conversions, propose a budget.
4. Preview and approval. Show the list of changes with current and new values, calculate the estimated change in daily spend and wait for my approval. After approval, apply ad group and keyword bids separately.

Set the priority order by [priority services, e.g. "motorcycle and heavy vehicle licences"]. Raising visibility for these comes before any other keyword.

OptimizationChanges after your approval

Stop paying to win customers you already have: exclude past converters

Confirms or builds a list of people who reached your thank-you page and excludes it from campaigns meant to win new customers. Never excludes a list based on its name alone when its definition cannot be read.

When to use itWhen you suspect the campaigns you pay to win new customers are also shown to existing customers, and your business does not depend on repeat purchases.

In my Google Ads account [account name], campaigns meant to win new customers may also be shown to people who already filled in a form or bought. I want to exclude these people from my acquisition campaigns. The address of my thank-you or order complete page is: [thank-you page URL]. My acquisition campaigns are: [campaign names, or write "you decide"].

1. Separate the campaigns. Get the enabled campaigns with account_snapshot. Leave out campaigns with brand in their name and remarketing campaigns. Leave out Performance Max campaigns too and note them separately, because this method does not work the same way there.
2. Read the existing lists. With list_audiences and definition=true, get every list with its definition, membership duration and Search and Display sizes. Never exclude a list whose definition cannot be read based on its name alone, because excluding the wrong list silently blocks real prospects too. In that case show me the list and ask me to confirm its rule in the Google Ads interface.
3. Pick or build the right list. If a list with a confirmed definition covers visitors of the thank-you page, use it. If not, propose a new one with create_website_audience, passing the thank-you page path as url_contains and [membership duration, e.g. 180 based on your sales cycle] as membership_days. Note that a new list takes time to fill and that Google needs at least 1,000 active members to use a list on the Search network; an exclusion with a smaller list has no effect on Search.
4. Read what is already attached. With google_ads_query, pull USER_LIST criteria with the negative field from campaign_criterion and ad_group_criterion. Do not propose adding the same list again where it is already excluded.
5. Preview and apply. For each campaign, use manage_audience_targeting with action=exclude and the list name or ID as audience; do not set bid_modifier. Show the preview for all campaigns with confirm=false first, wait for my approval, then apply and confirm by reading campaign_criterion again that the list was added with negative=true.

Exception: if my business runs on repeat purchases or renewals (subscription renewals, regular servicing, consumables), excluding these people may cut revenue. Ask me about this at the start, and depending on my answer propose a separate remarketing campaign instead of an exclusion.

Format the output like this: a campaign table (campaign, type, existing exclusions, list to add, list size on Search and Display, decision), a rough estimate of when the list may reach a usable size for Search, and the spend that went to this list in the last 30 days. You can measure that spend with google_ads_query from campaign_audience_view and ad_group_audience_view only if the list was already attached to the campaign in observation; if it was not, say it cannot be measured and do not guess.

OptimizationChanges after your approval

Send sales from Microsoft Ads clicks back as offline conversions

Uploads sales closed on the phone or in store to Microsoft Ads using the Microsoft Click ID (MSCLKID). Checks the offline goal and the data first, removes duplicate and incomplete rows and puts the preview up for your approval.

When to use itWhen your Microsoft Ads prospects turn into sales later on the phone, in store or in your CRM, and you want bidding to learn from real sales rather than form fills.

In my Microsoft Ads account [account name], some of the prospects that come from ads later turn into sales on the phone or in store. I want to send these sales back with the Microsoft Click ID (MSCLKID) so bidding learns from real sales. Here is my sales list: [paste the table with MSCLKID, sale date and time, time zone and sale amount columns].

1. Find the account. Get the account ID and currency with list_microsoft_accounts. If the currency comes back empty, do not guess, ask me.
2. Check the goals. With list_microsoft_conversion_goals, list the conversion goals in the account with their type, count type and conversion window (conversion_window_min, in minutes). If there is no goal of the offline type, stop and tell me. create_microsoft_conversion_goal only creates URL-based goals, so I need to create the offline goal myself in the conversion goals section of the Microsoft Advertising interface. Once I have, ask me for its exact name.
3. Clean the data. Check my rows against these rules:
   1. Remove rows with an empty MSCLKID and say how many there were.
   2. If the same MSCLKID appears twice, keep only one, because sending two sales for one click inflates conversions. If they really are two separate sales, ask me.
   3. Convert dates to UTC in the form 2026-08-01T14:30:00Z. If no time zone is given, do not guess, ask me.
   4. The sale must be after the click and inside the goal's conversion window. If the list has no click date, say plainly that this check could not be done.
   5. Turn amounts into numbers and compare the currency with the account currency.
4. Preview. Run upload_microsoft_offline_conversions with account_id, the goal's exact name as conversion_name, the cleaned rows as conversions (msclkid=...;datetime=...;value=..., one row per line) and the currency, with confirm=false first. Show the row count, a sample row and the total amount from the preview.
5. Send after my approval. Upload with the same parameters and confirm=true, and report the tool's result with the row count. Microsoft takes time to show conversions in reports, so remind me to check the next day with microsoft_report_query, pulling Conversions and Revenue from the CampaignPerformance report, whether the uploaded sales appear.

Finally, write a short checklist for a lasting setup: how to confirm in the Microsoft Advertising interface that auto-tagging of MSCLKID is on, how the site form should store the MSCLKID in a hidden field, and which columns the sales team must fill for each sale. Note that no tool checks these, so I will carry out these steps myself.

OptimizationChanges after your approval

Audit shared portfolio bid strategies and CPC limits

Lists the portfolio bid strategies in the account, which campaigns use each one and how actual results compare with the target. Finds unused portfolios, badly grouped campaigns and missing CPC limits, and puts the fixes up for your approval.

When to use itWhen your account carries shared bid strategies from the past, or when you want to cap what a single click can cost on a Smart Bidding campaign.

Audit the shared (portfolio) bid strategies and the CPC limits on campaigns in my Google Ads account [account name]. My target CPA is [target CPA] and my target ROAS is [target ROAS, or write "none"].

1. List the portfolios. With google_ads_query, pull only the strategies with bidding_strategy.status = 'ENABLED' from the bidding_strategy resource: name, type, non_removed_campaign_count, target CPA, target ROAS and the CPC ceiling and floor (for example maximize_conversions.cpc_bid_ceiling_micros, target_spend.cpc_bid_ceiling_micros). Removed strategies also come back from this resource, so do not skip the status filter. Strategies owned by a manager account do not appear here; pull them separately from accessible_bidding_strategy with owner_customer_id and owner_descriptive_name.
2. Map the campaigns. With google_ads_query, pull bidding_strategy_type, bidding_strategy and the campaign-level CPC ceiling (target_spend.cpc_bid_ceiling_micros) for every campaign that is not removed. Put into a table which campaign uses which portfolio and which one runs its own strategy.
3. Compare results. Pull the last 30 days at campaign level with get_report and compute total cost, conversions, cost per conversion and ROAS for each portfolio. Put the portfolio target next to the actual result.

Flag the following:
1. Enabled portfolios that no campaign uses.
2. Campaigns in the same portfolio that have different primary conversions or very different costs per conversion. A portfolio target applies to all its campaigns together, so brand and non-brand campaigns sharing a portfolio is usually a mistake.
3. Portfolios whose actual cost per conversion is above 1.5 times the target or below half of it. Do not judge a portfolio with fewer than 15 conversions in the last 30 days, write "not enough data".
4. Campaigns on Maximize conversions or Target CPA where I want to cap what a single click can cost. On these strategies a CPC ceiling cannot be set at campaign level, only through a portfolio strategy.
5. Campaigns on Maximize clicks without a CPC ceiling.

Format the output like this: first the portfolio table (portfolio, type, target, CPC ceiling, number of campaigns, actual cost per conversion, ROAS, finding), then the campaign table (campaign, strategy, portfolio, campaign CPC ceiling, finding), and finally a list of changes in priority order.

Make the changes with manage_portfolio_bidding_strategy: create or create_and_assign for a new portfolio, update to change the target or limit of an existing one, assign to attach a campaign to an existing portfolio, delete to remove a portfolio that no campaign uses. State clearly in the preview that changing a portfolio target affects EVERY campaign that uses it. Do not propose changing a target by more than 20% in one step. Preview every change with confirm=false first, wait for my approval, then apply it and confirm by reading the bidding_strategy and campaign resources again.