As of 13 August 2026, AI can only partly optimise your Google Ads.

What to actually do

1. Use a tool built for this

   The route this page recommends

2. Do it yourself

   Second choice

   A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open Google Ads and export the relevant campaign, ad group, keyword, search-term and performance reports for a recent period and a comparable earlier period.
   2. Gather your conversion definitions, tracking notes, budgets, bidding strategies, target cost or return, customer value or margin, UK targeting settings and any claims or brand restrictions.
   3. Paste the exports and business context into a chatbot with the prompt above, keeping campaign names and dates attached to each set of figures.
   4. Ask the model to produce the ranked plan, then remove any recommendation based on missing data, unsupported assumptions or a conversion metric you do not trust.
   5. Compare each proposed budget, bid, keyword, negative and advert change with the live Google Ads settings and your approved claims before applying anything.
   6. Apply only the approved changes in Google Ads, record the previous settings, and monitor the specified metrics and reversal conditions before making another change.

   Prompt

   Act as a paid-search analyst, not an autonomous account manager. Analyse the Google Ads information below and produce a cautious optimisation plan for a UK business.
   
   Business and goals:
   - Business: [describe the business and location]
   - Products or services: [describe them]
   - Primary conversion: [for example, completed purchase or qualified enquiry]
   - Target cost per conversion or return target: [insert target or say unknown]
   - Gross margin or value of a conversion: [insert figure or say unknown]
   - Brand restrictions, claims rules and excluded audiences: [insert details]
   
   Data and account context:
   - Date range and comparison range: [insert both]
   - Campaign type, bidding strategy and budget: [insert details]
   - Campaigns, ad groups, keywords, search terms and negatives: [paste export]
   - Impressions, clicks, cost, conversions, conversion value and other available metrics: [paste export]
   - Landing pages and relevant conversion-tracking notes: [paste details]
   
   Do not invent missing data, platform behaviour, search intent or business facts. Separate observed facts from hypotheses. First identify tracking or data-quality problems that could make optimisation unsafe. Then rank recommendations by expected impact and confidence, explaining the evidence for each. Cover budget allocation, search terms and negatives, keywords, bidding, ad strength and landing-page relevance, but only recommend a change where the supplied data supports it. For every recommendation include: the exact change, the reason, the main risk, what to monitor, and a clear stop or reversal condition. Draft replacement headlines and descriptions only from the supplied business facts, and flag every claim that needs approval. Finish with a short list of changes that should not be made until more information is available. Do not claim that any change will improve performance without a controlled test or subsequent evidence.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

3. Hand it to a person

   The distant third

   A person who owns the outcome does this end to end, worth it when the failure is dear.