Skip to main content
MethodMarketingPaid advertisingIntermediate

Audit search terms and negative keywords

A repeatable audit of Google Ads search terms, including AI Max and Performance Max: find wasted spend and new keyword ideas, group terms with n-gram analysis, and keep negatives in the right lists without blocking searches that convert.

Use this task 5 quick questions, then open it in ChatGPT, Claude or Gemini.

This is a practical, copy-paste process for you are doing weekly or monthly account maintenance. Use it when you want a clear structure, then add your own details.

Verified on 2026-10-228 sources cited

Quick start

1. Download your search terms report with cost and conversions, leaving out the last week or two if conversions arrive late.
2. Remove any names, phone numbers or emails before sharing it with an AI.
3. Look for searches that spent a lot and never converted, and for words that show up in many bad searches.
4. Before blocking a word, check it never appears in searches that did convert, and decide yourself.
5. If AI Max sent people to the wrong page, exclude the page instead of the search.
6. Add plurals and synonyms of each negative yourself.
7. Check the results 2 and 4 weeks later, then repeat every few weeks.

Before you start

Full method

How to use this method

  1. 1. Copy or downloadCopy the prompt below, or download the method file and attach it to your chat.
  2. 2. Paste into any AIChatGPT, Claude, Gemini or Copilot. Replace each [PLACEHOLDER] with your details, or let the AI ask you.
  3. 3. Add your materialAttach your own files, data or brief so the AI works from your facts, not guesses.

The prompt

Paste it as-is. The AI will ask for anything still in [BRACKETS].

Download .md
# Audit search terms and negative keywords
Method by Deepstory Research: deepstoryresearch.com/method/search-terms-and-negative-keyword-audit

You are a paid-search analyst who runs search-terms audits for Smart Bidding, broad match, AI Max and Performance Max accounts. Follow the method below step by step. Goal: A repeatable audit of Google Ads search terms, including AI Max and Performance Max: find wasted spend and new keyword ideas, group terms with n-gram analysis, and keep negatives in the right lists without blocking searches that convert.

## My inputs
- Search terms report: [SEARCH_TERMS_REPORT]  (example: attach CSV for the last 30 to 90 complete days, with personal data removed)
- Target cost per conversion or ROAS: [TARGET_COST_PER_CONVERSION_OR_ROAS]  (example: USD 60 per qualified lead)
- What you sell and do not sell: [WHAT_YOU_SELL_AND_DO_NOT_SELL]  (example: CRM software; no free plan, no jobs, no training)
- Conversion lag and primary conversion: [CONVERSION_LAG_AND_PRIMARY_CONVERSION]  (example: qualified lead, reported 7 to 14 days after the click)
- Campaigns in scope: [CAMPAIGNS_IN_SCOPE]  (example: all Search (some on AI Max) and Performance Max)

Before you start: if any input above is still in [BRACKETS] or missing, ask me for it in one short numbered list and wait. Do not guess. If I attached files (data, briefs, Deepstory evidence files), treat them as the primary source.

## Steps to follow
1. Check the period excludes the latest conversion-lag days, the conversion columns use the primary goal, and how much spend the report does not show.
2. Separate AI Max matches (match type and Source columns) and Performance Max terms from keyword matches.
3. Tag each term keep, watch, add as keyword, or negative candidate, using the thresholds below.
4. Run an n-gram analysis to find words and phrases that waste money across many terms.
5. Check every candidate against converting terms and the business offer, and confirm the list with the user.
6. Pick the narrowest level and the right tool for each fix: negative keyword, shared list, URL exclusion or brand exclusion.
7. Add plurals and synonyms that negatives will not catch on their own.
8. Log the changes, plan the 14 and 28-day check and the next audit date.

## Deliver the output in this structure
- Report check (period, lag, hidden spend)
- Term decision table
- N-gram waste table
- Proposed fixes by level and tool
- Conflict check
- New keyword candidates
- Change log with follow-up dates

## Quality rules
- Never propose a negative that would block a term that converted without saying so.
- Default negatives such as "free" or "jobs" are only candidates until the user confirms them.
- Label waste thresholds and cadences as rules of thumb.
- Label every number as FACT (with source and date), ESTIMATE (show the calculation) or ASSUMPTION.
- If a Deepstory evidence file is attached, use its figures as the reference values and cite its source references; tag claims that are not from it as [UNVERIFIED].
- End with: the 3 assumptions that matter most, what would change the answer, and the single best next step.

What you need to give the AI

InputPlaceholderExample
Search terms report[SEARCH_TERMS_REPORT]attach CSV for the last 30 to 90 complete days, with personal data removed
Target cost per conversion or ROAS[TARGET_COST_PER_CONVERSION_OR_ROAS]USD 60 per qualified lead
What you sell and do not sell[WHAT_YOU_SELL_AND_DO_NOT_SELL]CRM software; no free plan, no jobs, no training
Conversion lag and primary conversion[CONVERSION_LAG_AND_PRIMARY_CONVERSION]qualified lead, reported 7 to 14 days after the click
Campaigns in scope[CAMPAIGNS_IN_SCOPE]all Search (some on AI Max) and Performance Max

Use it when

  • You are doing weekly or monthly account maintenance.
  • Broad match or AI Max is matching to searches you do not want.
  • Cost per conversion rose and you suspect irrelevant searches.

Best with: A chat AI that can read CSV files. Remove names, phone numbers and emails from the CSV before you attach it.

What you get

  • Report check (period, lag, hidden spend)
  • Term decision table
  • N-gram waste table
  • Proposed fixes by level and tool
  • Conflict check
  • New keyword candidates
  • Change log with follow-up dates

The AI will ask you first

Answer these before the AI advises you; each one changes what it recommends.

  1. 1

    What does the business sell, to whom, and where?

    Relevance can only be judged against the real offer, audience and locations.

    If unknown: Ask; do not judge relevance without it.

  2. 2

    What do you not sell or not serve (free versions, jobs, DIY, courses, other brands, other places)?

    These are the fastest negative candidates, but the same words are core intent for some businesses (recruiters, trainers, DIY shops, free-tier software).

    If unknown: Assume none. Propose free, jobs, careers, salary, course, DIY and how to as candidates only, for the user to confirm.

  3. 3

    Which conversion is primary, does it measure sales or qualified leads, and how long after the click is it usually recorded?

    Terms with late or lower-quality conversions are easy to cut by mistake.

    If unknown: Exclude the last 7 days, or 14 days for B2B leads, from the judging period (rule of thumb).

  4. 4

    What is the target cost per conversion, or the ROAS you need?

    Waste is judged against what a conversion is worth to you.

    If unknown: Use the account average for the period.

  5. 5

    Which bidding strategies and match types are used, and which campaigns run AI Max or Performance Max?

    AI Max and Performance Max terms need their own handling, and new AI Max campaigns need time before negatives.

    If unknown: Assume broad match with Smart Bidding and check the Match type column for "AI Max".

  6. 6

    What negative lists exist today, which campaigns use them, and do you have brand and non-brand campaigns?

    New negatives should go into the right list, and a shared list must never block your own brand campaign.

    If unknown: Assume only campaign-level negatives exist.

The procedure, step by step

Read this before you run it, so you can check the AI actually followed the method.

  1. 1

    Export the report

    Campaigns > Insights and reports > Search terms. Add Match type and Source columns. Use 30 complete days ending after the conversion lag, or 60 to 90 days for accounts under about 30 conversions a month (rule of thumb). Remove personal data from the CSV.

  2. 2

    Check coverage

    Report cost vs campaign cost = hidden share; use search terms insights for the hidden themes. Note AI Max and Performance Max terms separately.

  3. 3

    Tag each term

    Keep, watch, add as keyword, or negative candidate, using the thresholds.

  4. 4

    Run n-grams

    Split terms into 1, 2 and 3-word pieces. Sum cost and conversions per piece. Sort by cost with zero conversions.

  5. 5

    Check conflicts and confirm

    Test each candidate against converting terms, keywords and the brand campaign. The user confirms every negative before it goes on a list.

  6. 6

    Pick level and tool

    Account level for never-relevant words; shared list for groups; campaign or ad group to route searches; URL exclusion for bad AI Max landing pages; brand lists and brand exclusions for brand problems.

  7. 7

    Add variants

    Plurals and synonyms of each negative. Avoid symbols such as ! @ % in negatives.

  8. 8

    Log and follow up

    Record what changed and why. Check volume, conversions and cost per conversion 14 and 28 days later; remove a negative if converting volume drops. Next audit weekly for new broad or AI Max campaigns, then every 2 to 4 weeks (rule of thumb).

Platform settings(show)
PlatformPathSettingNote
Google AdsCampaigns > Insights and reports > Search termsSearch terms report (download as CSV)
Google AdsSearch terms report > columns: Match type, Source; view: Search terms and ad combinationsAI Max search terms and landing pages; "Add as negative URL"
Google AdsTools > Shared library > Brand listsBrand lists for AI Max brand settings
Google AdsTools > Shared library > Exclusion lists (negative keyword lists)Shared negative keyword listsUp to 20 lists, 5,000 keywords each. UI path as of October 2026.
Google AdsAdmin > Account settings > Negative keywordsAccount-level negativesUp to 1,000; includes Performance Max Search and Shopping inventory.
Google AdsCampaigns > Keywords > Negative keywords > + > Performance Max campaignPerformance Max campaign negatives or list

Worked example

The figures below are illustrative. Use your own numbers and facts.

Scenario: An Indian B2B SaaS company running broad-match Search (two campaigns on AI Max since August) and Performance Max for demo requests; target USD 60 per qualified lead; leads qualify 7 to 14 days after the click.

Input: 60-day report ending 14 days ago: 3,100 terms, USD 15,800 spend (report covers USD 12,900, so about 18% hidden), 240 leads, 96 qualified. No shared lists; a few campaign negatives.

Output excerpt: N-grams with zero qualified leads: "course" (USD 520 across 31 terms), "jobs" (USD 210), "open source" (USD 640). "free trial" produced 9 qualified leads, so "free" is not a broad negative; propose phrase negatives "free download" and "free version" for the user to confirm. AI Max terms sent "crm pricing india" to the blog; propose "Add as negative URL" for /blog instead of a keyword negative. New shared list "Non-buyer intent" for non-brand campaigns only: jobs, course, open source plus plurals; salary added only if the owner confirms. Check results on 5 and 19 November.

Check your result

Before you trust the AI's output, confirm each of these.

  • ☐ The period excludes the latest conversion-lag days.
  • ☐ AI Max and Performance Max terms were reviewed separately.
  • ☐ Every negative was checked against converting terms and confirmed by the user.
  • ☐ Each fix uses the narrowest level and the right tool.
  • ☐ Plurals and synonyms were added.
  • ☐ A dated change log exists with 14 and 28-day follow-up checks.
What you'll get(show template)
## Search terms audit: [ACCOUNT] ([date range], ending [lag] days ago)

Spend in report: [amount] of [campaign total] ([x]% hidden) | Conversions: [n, primary goal] | Target: [CPA or ROAS]

### Term decisions
Search term | Match type / Source | Cost | Conv. | Decision | Reason

### N-gram waste
N-gram | Terms | Cost | Conv. | Proposed action

### Proposed fixes (user to confirm)
Tool (negative / shared list / URL exclusion / brand list) | Level | Item | Variants | Applies to

### Conflicts checked
Candidate | Converting term it could block | Resolution

### New keywords
Keyword | Match type | Ad group | Evidence

### Change log
Date | Change | Why | Check on (14 days) | Check on (28 days)

Quality checks built into the prompt

  • • Never propose a negative that would block a term that converted without saying so.
  • • Default negatives such as "free" or "jobs" are only candidates until the user confirms them.
  • • Label waste thresholds and cadences as rules of thumb.
  • • Every number labelled FACT, ESTIMATE or ASSUMPTION; claims outside attached evidence tagged [UNVERIFIED].

Decision rules

If/then rules the AI should apply. Heuristic rules are a practitioner's rule of thumb, not a platform rule.

IfThenType
You read the search terms report.Low-volume searches are hidden for privacy, and the hidden share can be large. Compare the report's total cost with campaign cost to see how much spend is unexplained.Platform rule
A campaign uses AI Max for Search.Add the Match type and Source columns and review "AI Max" matches separately. Use the "Search terms and ad combinations" view to see which landing page each term got, and fix a bad page with "Add as negative URL" rather than a keyword negative.Platform rule
AI Max was turned on less than 2 weeks ago.Wait before adding negatives, except clearly irrelevant or brand-unsafe terms; Google advises at least 2 weeks and using negatives sparingly.Threshold: 2 weeksPlatform rule
AI Max matches competitor or unwanted brand searches.Use a brand list (Tools > Shared library > Brand lists) applied in the campaign's AI Max settings, instead of many keyword negatives.Platform rule
You add a negative keyword.Add plurals, synonyms and close variants separately; negatives do not match them. Words after the 16th word of a search are ignored.Platform rule
You choose a negative match type.Broad negative blocks searches containing all its words in any order; phrase blocks the words in order; exact blocks only that exact search. Use exact negatives to route one search between your own campaigns.Platform rule
A term should never trigger a Search or Shopping ad in any campaign.Use the account-level negative list (up to 1,000 terms); it applies to Search and Shopping inventory in all campaign types, including Performance Max.Platform rule
A negative applies to a group of campaigns.Put it in a shared negative keyword list (20 lists of up to 5,000 terms each), and check the list is not attached to your brand campaign.Platform rule
You audit Performance Max.Add campaign negatives or attach a list (up to 10,000 per campaign, per the Google Ads API criteria doc); they affect Search and Shopping only. For Performance Max serving on your own brand, use brand exclusions rather than negating your brand name.Platform rule
You negate a term in one campaign.Check where it serves afterwards: it can move into Performance Max, AI Max or another broad campaign. Look for it again in the next report and add it there too if needed.Heuristic
The account has campaigns that used automatically created assets or campaign-level broad match (auto-upgraded to AI Max from September 2026), or still runs Dynamic Search Ads.Expect AI Max matches in the report for the upgraded campaigns. Dynamic Search Ads auto-upgrade from February 2027 and keep their URL controls, so audit their negatives and URL exclusions now.Platform rule
A term spent more than about 2 times target cost per conversion with no conversions, measured on data that excludes the latest conversion-lag window.Make it a negative candidate, after the conflict check. If qualified-lead or offline conversions are imported, judge on those, not raw form fills.Threshold: about 2x target CPA with 0 conversions (rule of thumb)Heuristic
The account bids on value (Target ROAS or Maximize conversion value).Make a term a candidate when its cost exceeds about 2 times (average order value / target ROAS) with no value, or its ROAS stays below about half of target over 30 or more days and 3 or more conversions.Threshold: rule of thumbHeuristic
An n-gram appears in many terms that together spent more than about 3 times target CPA with no conversions.Propose it as a broad or phrase negative at the narrowest level that covers it.Threshold: about 3x target CPA across terms (rule of thumb)Heuristic
A term has 2 or more conversions at or below target over 30 or more days and is not a keyword.Add it as an exact or phrase keyword where the landing page fits, mainly for control of copy and landing page.Threshold: 2 or more conversions (rule of thumb)Heuristic

Failure modes

Common mistakes with this method, how to spot them, and how to fix them.

MistakeSymptomFix
Adding "free" as a broad negative when "free trial" converts.Lead volume drops after the audit.Run the conflict check and use narrower phrase negatives.
Using default negatives without asking.A recruiter or trainer blocks its own best searches.Treat defaults as candidates the user confirms.
Judging B2B terms on the last 7 days.Good terms cut before their leads were recorded.Exclude the conversion-lag window from the period.
Adding many negatives in the first week of AI Max.AI Max cannot learn and volume stalls.Wait about 2 weeks, then use negatives sparingly.
Blocking a keyword when the real problem is the landing page AI Max chose.Good searches lost.Use a URL exclusion instead.
A shared negative list attached to the brand campaign.Brand searches stop showing ads.Check list attachments after every change.
Negating a term in one campaign only.The same search moves into Performance Max, AI Max or another broad campaign.Check the next report across all campaigns and add the negative where it reappears.
Assuming the report shows every search.Hidden waste stays unexplained.Measure the hidden share and use n-grams and search terms insights.

Agent mode: if an AI agent runs this unattended

May do without asking

  • • Analyse a search terms CSV that the user exports, with personal data removed, and attaches.
  • • Run n-gram analysis and draft negative, URL exclusion and keyword proposals.
  • • Check proposals against converting terms.

Must stop before (the user decides and does these)

  • • Adding, removing or attaching negatives, keywords, lists, URL exclusions or brand lists in the live account, which changes spend.
  • • Posting from Google Ads Editor, or applying or auto-applying recommendations.
  • • Publishing or enabling any campaign.
  • • Uploading files to the ad account.
  • • Accepting platform terms or changing billing.

Evidence for this method

This method runs on your own material: your files, brief or code. If your task needs market or economic numbers, attach a sourced context file from the evidence library alongside the prompt.

Browse evidence

Run next

Methods that chain well after this one.

Deepstory methods are process guidance for AI tools. Review AI output before relying on it; legal, tax, medical and investment decisions need a qualified professional.