How to Do Keyword Research and Build a Negative Keyword List — closing the leak Smart Bidding can't see

Keyword Planner tells you what queries *might* exist and roughly how often; the Search Terms report tells you what queries actually triggered your ads. Research without pruning wastes budget on irrelevant traffic — pruning without research starves the campaign of reach. This page gives the method for both sides and the exact match-type trap that breaks negative keywords silently.

Using Keyword Planner for Volume/Competition Estimates and the Limits of Its Data

Keyword Planner is a tool within Google Ads that helps advertisers discover keywords related to their products, services, or websites, and returns estimated monthly search volumes, average cost-per-click data, and competitive metrics — FONTE: https://support.google.com/google-ads/answer/7337243

Treat its numbers as directional, not predictive:

- Forecasts are based on a full week of historical search data, averaged to produce a daily estimate, adjusted for seasonality, and may not account for unexpected market shifts — FONTE: https://support.google.com/google-ads/answer/3022575 - Keywords with very low search volumes are not forecastable — Keyword Planner returns an error asking you to try new keywords instead of a number — FONTE: https://support.google.com/google-ads/answer/7337243 - Sensitive categories (e.g. health, finance) are not discoverable or forecastable in some cases — FONTE: https://support.google.com/google-ads/answer/7337243 - Actual campaign performance depends on factors the forecast doesn't model: bid amount, budget, ad quality, location targeting, industry trends, Display Network activity, cross-campaign competition, and timing — FONTE: https://support.google.com/google-ads/answer/3022575 - New advertiser accounts get forecasts built on historical *average* data across all advertisers, not account-specific data, because there isn't enough account history yet — FONTE: https://support.google.com/google-ads/answer/3022575

When it applies

use Keyword Planner to seed a first keyword list and rough-size a market before launch. Do not use it to promise a client a specific volume or CPC — the tool itself documents that real results diverge. This matters for Bidding Strategies: Manual, Smart Bidding, and When Each Applies — matching the strategy to the conversion data you actually have: a bidding strategy set on a Keyword Planner forecast alone has no real conversion signal behind it yet.

Mining the Search Terms Report as an Ongoing (Not One-Time) Research Source

The Search Terms report displays the actual queries that a significant number of people searched for which triggered your ads, together with performance metrics — clicks, impressions, conversions, cost — per query — FONTE: https://support.google.com/google-ads/answer/2472708

This is the correction mechanism Keyword Planner can't provide: it shows what really happened, not what might happen. Two uses:

- Find waste: identify searches related to, but not exactly, what you sell, and add them as negative keywords — FONTE: https://support.google.com/google-ads/answer/7102466 - Find opportunity: discover new keyword ideas and content ideas by seeing what customers are actually searching for — FONTE: https://support.google.com/google-ads/answer/2472708

Google omits low-volume queries from the report for privacy reasons, though search-terms insights may aggregate these into themes — so the report is not a complete query log, and you won't see everything a small-volume query triggered — FONTE: https://support.google.com/google-ads/answer/2472708

This is not a one-time audit step. Regular monitoring of the Search Terms report is necessary to continuously refine targeting and reduce wasted ad spend — FONTE: https://support.google.com/google-ads/answer/7102466. See the cadence section below for how often without over-pruning.

Building a Negative Keyword List: What to Exclude and Where

Negative keywords are terms that prevent your ads from showing when users search for those terms, letting you focus campaigns on relevant traffic and improving ROI by only showing ads to interested users — FONTE: https://support.google.com/google-ads/answer/2453972

Mechanics to know before building the list:

- You can create up to 20 negative keyword lists per account, each holding up to 5,000 negative keywords — FONTE: https://support.google.com/google-ads/answer/6372658 - Negative keywords can be added at three levels: ad group, campaign, or a shared negative keyword list — FONTE: https://support.google.com/google-ads/answer/7102466 - By default, negatives added directly from the Search Terms report are set as negative exact match — FONTE: https://support.google.com/google-ads/answer/7102466 - Shared lists don't work until linked: a negative keyword list created in the Shared Library must be explicitly applied to campaigns — if it's created but not associated, it excludes nothing — FONTE: https://support.google.com/google-ads/answer/9701952 - Overdoing it backfires: using too many negative keywords can reduce the number of customers your ads reach — FONTE: https://support.google.com/google-ads/answer/2453972 (expanded on in the cadence section)

The notes for this page name three recurring categories worth a standing shared list: irrelevant-intent terms, competitor-name confusion, and job-seeker/free-seeker terms (e.g. "jobs", "salary", "free", "diy") — these are the categories Google's own guidance points to when it says to exclude searches "related to, but not exactly, what you sell" — FONTE: https://support.google.com/google-ads/answer/7102466. Put this shared list to work across every campaign for the brand: it's the natural pairing with How to Structure a Search Campaign (Ad Groups, RSAs, Settings) — the foundation an audit checks and a budget plan assumes is sound, which places negative keyword lists at the campaign vs. shared account level.

Match Type Choice for Negatives and the Broad-Match Negative Conflict Trap

For Search campaigns, negatives come in the same three flavors as positive keywords — negative broad match, negative phrase match, negative exact match — but they behave differently from their positive counterparts — FONTE: https://support.google.com/google-ads/answer/2453972

- Negative broad match: blocks a search only if it contains ALL words of your negative phrase, in any order. If the search contains only *some* of the terms, the ad can still show — FONTE: https://support.google.com/google-ads/answer/7302703 - Negative phrase match: blocks a search that contains your exact phrase in the same sequence — FONTE: https://support.google.com/google-ads/answer/2453972 - Negative exact match: blocks a search only when the query is precisely your keyword — FONTE: https://support.google.com/google-ads/answer/2453972 - Negatives do not expand: unlike positive keywords, negative keywords do NOT match to close variants, misspellings, reordered words, or other expansions — FONTE: https://support.google.com/google-ads/answer/2453972. This is the piece that trips people up most — see Keyword Match Types in 2026: Broad, Phrase, Exact — and Why None of Them Is Literal Anymore for how positive match types now expand via close variants, and note this is the opposite behavior for negatives.

The conflict trap

an ad group cannot target and exclude the same keyword — if a negative keyword conflicts with a positive keyword in the same ad group, Google Ads Editor flags it as a conflicting target/exclusion error — FONTE: https://support.google.com/google-ads/editor/answer/79781. This is the concrete failure mode of What Does NOT Work: Broad Match and Smart Bidding Without Guardrails — where automation burns budget instead of saving it: a broad negative added carelessly at the campaign or shared-list level can silently block a positive keyword you're actively bidding on elsewhere in the account, and because negatives don't expand, a near-miss spelling of the negative won't be caught either.

Also watch query length

searches exceeding 16 words may still trigger ads even if a negative keyword appears after the 16th word in the query — a query truncation issue independent of match type — FONTE: https://support.google.com/google-ads/answer/9701952

Cadence: How Often to Review Without Over-Pruning and Starving Smart Bidding of Data

Google does not prescribe a fixed schedule (e.g. "weekly" or "monthly"). Instead, the guidance is to focus on finding negative keywords that occur with reliable frequency in the Search Terms report, rather than reading through every single query — FONTE: https://support.google.com/google-ads/answer/7102466

Practical reading of that guidance:

- Look for recurring themes in queries that aren't showing strong ROI, not one-off low-performing searches — FONTE: https://support.google.com/google-ads/answer/7102466 - Push confirmed, durable exclusions (terms that will never fit the brand) into the shared library so they apply across all campaigns automatically, rather than re-adding them campaign by campaign — FONTE: https://support.google.com/google-ads/answer/7102466 - Don't over-prune. Excessive use of negative keywords can reduce campaign reach and, by extension, the volume of traffic and conversion data available for Smart Bidding to optimize against — FONTE: https://support.google.com/google-ads/answer/2453972. A related principle applies to data exclusions in Smart Bidding — those aren't intended to be used frequently or for extended periods, since doing so can hurt Smart Bidding performance — FONTE: https://support.google.com/google-ads/answer/10276486 *(unverified as directly equivalent: this source addresses conversion-data exclusions, not negative keywords, but the notes flag it as the same underlying risk — thinning the data Smart Bidding learns from)*.

Why this page serves the wiki's objective

the objective is putting a marketer or their AI agent in a position to run campaigns without burning budget. Keyword research and negative keywords are the most direct lever against silent budget waste — irrelevant clicks that never convert — but the match-type and conflict-trap details above are exactly where a naive automation (add every Search Terms report query as a broad negative, on a schedule) would break a campaign by blocking its own positive keywords or thinning the data Bidding Strategies: Manual, Smart Bidding, and When Each Applies — matching the strategy to the conversion data you actually have needs to exit its learning period.

What does NOT work

- Treating Keyword Planner volume as a guarantee: the tool's own documentation says new-account forecasts use cross-advertiser averages and actual results depend on many factors it doesn't model — FONTE: https://support.google.com/google-ads/answer/3022575 - Adding negatives as broad match by default expecting them to behave like a catch-all: negative broad still lets ads show if only *some* of the words are present, and won't catch misspellings or reordering the way you might expect from positive broad match — FONTE: https://support.google.com/google-ads/answer/7302703, https://support.google.com/google-ads/answer/2453972 - Creating a shared negative keyword list and assuming it's active: if it isn't linked to campaigns, it excludes nothing — FONTE: https://support.google.com/google-ads/answer/9701952 - Reviewing every single query line by line on a fixed schedule: Google's own guidance says to focus on recurring-frequency terms instead — FONTE: https://support.google.com/google-ads/answer/7102466 - Aggressive, frequent negative-keyword pruning as a way to "clean up" a campaign: it can shrink reach and starve Smart Bidding of the data volume it needs — FONTE: https://support.google.com/google-ads/answer/2453972

Related

- Keyword Match Types in 2026: Broad, Phrase, Exact — and Why None of Them Is Literal Anymore — positive match types expand to close variants; negatives explicitly do not. Read both pages together before touching an existing keyword list, or you'll misjudge what a negative actually blocks. - How to Structure a Search Campaign (Ad Groups, RSAs, Settings) — the foundation an audit checks and a budget plan assumes is sound — decides whether a negative list lives at the campaign level or the shared account level; this page supplies the content and cadence for that list. - What Does NOT Work: Broad Match and Smart Bidding Without Guardrails — where automation burns budget instead of saving it — the failure mode when negative keywords (or their absence) interact badly with Smart Bidding's need for conversion data; this page's cadence section is the preventive practice. - How to Audit an Existing Google Ads Account (Priority Order) — the sequence that stops you from trusting broken data — an account audit's step reading the Search Terms report is exactly the research method described here, applied retroactively to an existing account.