Google Ads Account Limits 2026: Every Number, and Where Structure Breaks Long Before the Ceiling
Limits usually surface at the worst moment: a bulk upload from Editor fails with an entity size error, a script dies halfway through, and the web interface explains nothing. But Google Ads account limits are not trivia. They define what structure you can build at all, and — more usefully — they tell you where a growing account will start to fall apart.
Below is the complete 2026 reference, plus the part documentation never covers: which limits you will actually hit, which you never will, where account performance degrades long before any formal ceiling, and how to design structure so you are not rebuilding it on live budgets.
Google Ads account limits: the full table
Campaigns and ad groups
| Entity | Limit | Notes |
|---|---|---|
| Campaigns per account | 10,000 | active plus paused |
| Ad groups per campaign | 20,000 | Local and App campaigns: 100 |
| Targeting items per ad group | 20,000 | keywords, placements, audiences combined |
| Ad group targeting items per account | 5,000,000 | the real ceiling for large search accounts |
| Campaign targeting items per account | 1,000,000 | geo, devices, scheduling and similar |
Ads and assets
| Entity | Limit | Notes |
|---|---|---|
| Text ads per ad group | 50 active | |
| Responsive search ads per ad group | 3 enabled | hard cap, no workaround |
| Image ads per ad group | 300 | |
| Ads per account | 4,000,000 | active plus paused |
| Ad group level assets per account | 250,000 | |
| Campaign level assets per account | 50,000 | |
| Customizer attributes for RSAs | 40 | 90 character limit per value |
Negatives and exclusions
| Entity | Limit | Notes |
|---|---|---|
| Negative keywords per campaign | 10,000 | Display and Video: 1,000 |
| Negative keyword lists per manager account | 20 | the most painful limit in practice |
| Keywords per negative list | 5,000 | |
| Shared placement exclusion lists per account | 20 | |
| Account level placement exclusions | 65,000 | |
| Total exclusions per ad group | 128,000 |
Geography and account management
| Entity | Limit | Notes |
|---|---|---|
| Location targets per campaign | 10,500 | 10,000 standard plus 500 proximity |
| Dynamic Search Ad targets per account | 25,000 | relevant during AI Max migration |
| Non-manager accounts per MCC | 85,000 | |
| Account creation rate | 2,500 per week | a rate limit, not a volume limit |
Which limits you will actually hit
The numbers above read like infinite headroom. In reality accounts collide with a very short list of constraints, and almost never with the big ones.
The three you will hit repeatedly:
- Three enabled RSAs per ad group. This is the binding constraint for anyone who tests creative at the ad level. Two ways out: test inside the RSA using assets and the asset performance report, or split variants across ad groups — at the cost of fragmenting conversion data across them. Mechanics in Ad Strength and the RSA asset report.
- Twenty negative keyword lists per manager account. This looks like nothing until you manage a dozen clients or verticals, each wanting its own set. Twenty runs out fast. The fix is not more lists but fewer, larger, universal ones — junk, irrelevant geos, competitors, informational queries — with campaign-level negatives handling specifics. Methodology in negative keywords and search term analysis.
- Twenty shared placement exclusion lists. Same shape of problem, same answer: one large list rather than fifteen thematic ones. See placement exclusions and brand safety.
The ones you will never hit: 10,000 campaigns, 4 million ads, 85,000 accounts under one MCC. If you are approaching those, the limit is not the problem — the structure stopped being manageable a long time ago.
The real ceiling arrives far earlier
This is the point worth taking away. Formal limits describe when the API returns an error. They say nothing about when the account stops functioning well, and that happens much sooner.
The interface degrades around 3,000–5,000 campaigns
Treat that as a working estimate rather than a published threshold, but it reproduces reliably: at that size reports take tens of seconds, filters lag, and bulk edits through the web UI become a gamble. The account is still workable — through Google Ads Editor and bulk operations, not through the browser.
Smart Bidding breaks on data, not on entity counts
A far harder constraint than any documented limit: conversion volume per campaign. The widely used benchmark is roughly 30 conversions in 30 days per strategy for learning to mean anything — treat it as an orientation point, not a guarantee. Which means the number of campaigns your account can carry is really your monthly conversion volume divided by that figure, not 10,000.
An account running 200 campaigns on 300 monthly conversions performs worse than the same budget in 15 campaigns. The binding limit here is statistical, and it arrives three orders of magnitude earlier than the technical one.
The practical consequence: splitting structure “for control” is nearly always paid for in bidding quality. What happens during learning and how long it lasts is covered in the Smart Bidding learning period.
Scripts hit execution time, not entity limits
Google Ads scripts carry their own execution ceiling (around 30 minutes for a standard run). On a large account, a script that iterates every campaign and keyword never finishes, regardless of how far you are from any entity limit. The rule that follows: at scale, automation is built on reports and batches, not on object iteration. When scripts stop being enough, the comparison is in Google Ads API versus scripts.
Designing structure that scales
Rule 1: account-level limits matter more than campaign-level ones
Note the asymmetry — 5 million ad group targeting items per account versus 1 million campaign targeting items. A five-to-one ratio in favour of ad group level targeting. If your structure leans on many campaigns with elaborate geo settings and lists, you will exhaust the campaign million long before the ad group five.
Practical read: push shared settings up to account level and shared lists, keep granularity in ad groups. It also shrinks the edit surface every time something changes.
Rule 2: one campaign, one budget decision
The most common source of unmanageable accounts is splitting campaigns along dimensions that do not need separate budgets. A separate campaign is justified when you need separate budget control, separate geography, separate scheduling or a separate bid strategy. Everything else is an ad group. Full method in Google Ads account structure for scale.
Rule 3: count assets before automation does
250,000 ad group assets sounds generous — right up until automatic generation is switched on. Auto-created assets, extensions, PMax image variants and video assets accumulate faster than expected, and unlike keywords nobody ever prunes them. Once a quarter, review zero-impression assets and delete. What belongs in assets at all is covered in ad extensions and assets.
Rule 4: design MCC hierarchy around people, not accounts
85,000 accounts per manager is a number you will never approach. The real constraint is access control and auditability: who can change what, and how fast you can establish who broke something. Hierarchy should mirror responsibility, not the client list. Mechanics in manager account and MCC operations.
The limits nobody lists
Beyond entity counts there is a second layer: limits on content. They are smaller, and they are the usual reason bulk uploads fail.
Character limits
| Field | Limit |
|---|---|
| RSA headline (up to 15) | 30 characters |
| RSA description (up to 4) | 90 characters |
| Display path | 15 characters per segment, two segments |
| Customizer value | 90 characters |
| Sitelink text | 25 characters; descriptions 35 each |
| Campaign, ad group and list names | typically 255 characters |
The trap: limits are measured after customizer substitution, not before. A headline that validates at 28 characters in your spreadsheet can fail at upload because the inserted value pushes it over. Mechanics in ad customizers and business data feeds.
Bulk operation limits
This is where most “entity exceeds size limits” errors actually live. Working guidance:
- keep API mutations to roughly 5,000 operations per request — not because more is forbidden, but because one bad operation makes a larger batch painful to debug;
- scripts carry their own cap on entities modified per run, and it is lower than expected; split large edits across scheduled runs;
- Google Ads Editor is constrained by your machine’s memory rather than by the platform — on accounts with hundreds of thousands of keywords, downloading everything is the mistake. Pull only the campaigns you are working on.
Limits imported from adjacent systems
Some ceilings are not Google Ads ceilings at all but hurt just as much: product counts and refresh frequency in the feed, audience limits in GA4, BigQuery export quotas, upload size caps for offline conversions. When the account looks healthy but the data is thin, the constraint is usually upstream. See Google Ads BigQuery Data Transfer and reporting.
What limits look like in three account types
Ecommerce with a large feed
Asset headroom runs out first: PMax asset groups, auto-generated images and video, dynamic extensions. Keyword counts may be near zero. The bottleneck is not 5 million targeting items but 250,000 assets and the discipline to prune them. The second constraint is feed quality and refresh rate rather than the ad account itself.
Search lead generation
Here the spend is in ad group targeting items and negatives. An account that has been harvesting negatives from the search terms report for years approaches 10,000 per campaign without anyone noticing. In parallel it collides with three RSAs per ad group, because all the work happens in copy. The fix is campaign consolidation plus moving testing inside the ad.
Agency account across many clients
Technical limits barely register, with two exceptions: 20 negative keyword lists and 20 shared placement exclusion lists per manager account. Those make “one list per client” impossible. The real constraint is not the platform but access hygiene and how fast you can audit a change.
What to do when you do hit a limit
| Limit | Wrong move | Right move |
|---|---|---|
| 3 RSAs per ad group | cloning ad groups | test at asset level inside the RSA |
| 20 negative keyword lists | requesting a second MCC | consolidate into 4–6 large lists |
| 10,000 negatives per campaign | moving some to another campaign | prune — at that volume half no longer appear in queries |
| 20,000 targeting items per ad group | splitting into ten ad groups | ask why one ad group holds 20,000 keywords; it is almost always a symptom |
| 250,000 assets | disabling automation entirely | scheduled pruning of zero-impression assets |
| 2,500 accounts per week | creating through a second MCC | revisit the operating model; that rate usually signals a process problem |
On the “entity exceeds size limits” error specifically: it comes from the API and usually means a single operation or object is oversized — an over-long list, too many items in one mutation — not that the account has hit a global cap. Reduce batch size before hunting for a system limit.
Four misconceptions about limits
“Limits only matter for huge accounts.” The three most painful constraints — three RSAs per ad group, twenty negative keyword lists, twenty placement exclusion lists — are entirely size-independent. An agency with ten clients hits them, and so does a 50-campaign account.
“More campaigns means tighter control.” Up to a point, yes. Past it, every additional campaign takes conversions from its neighbours and degrades learning across all of them simultaneously. Control bought with statistics is illusory: you end up steering campaigns that have nothing left to learn from.
“Pausing is safer than removing, so leave it.” Pausing does preserve history, but paused entities consume limit headroom and slow the interface. The rule: pause what might genuinely return, remove the rest deliberately.
“An entity size error means the account is full.” Almost never. In the overwhelming majority of cases it is one oversized operation in the upload. Shrink the batch first; only if the error repeats should you go looking for a system ceiling.
Quarterly hygiene
- Paused entities. They consume limit headroom exactly like active ones. A campaign that has not run in a year should be removed, not paused.
- Keywords with zero impressions over 90 days. In large accounts this is routinely half the list.
- Negatives that never blocked anything. They accumulate for years and never get reviewed.
- Zero-impression assets, particularly auto-generated ones.
- Duplicate lists. Two lists overlapping 80% are one list and one wasted slot out of twenty.
- Structure against conversions. Divide monthly conversions by campaign count. Under 30, the structure is too fragmented regardless of any limit.
If the account is already large and restructuring feels risky, the order stays the same: prune, then consolidate, then grow entity counts. To pressure-test a specific structure and find its real ceiling, see Google Ads training. If you need an account built for scale from day one, that is Google Ads agency accounts.
Related reading
Limits are the technical ceiling. There are two others: the policy ceiling, which decides which audiences exist for you at all — see the Google Ads personalized advertising policy — and the creative ceiling, covered in creator partnerships boost and YouTube creator videos in ads.
FAQ
Do paused campaigns count toward limits?
Yes. The 10,000 campaign and 4 million ad limits count active and paused together. Removed entities do not count.
Why can’t I add a fourth RSA to an ad group?
Three enabled responsive search ads per ad group is a hard format cap with no workaround. It is deliberate: the system is designed for testing inside an ad through assets, not across ads.
What if I need more than 20 negative keyword lists?
Consolidate. In almost every case, twenty lists are five meaningful sets smeared across themes. Build large universal lists and use campaign-level negatives for specifics.
Is the 5 million targeting item limit just keywords?
No. It covers every ad group level targeting item: keywords, placements, audiences, topics, demographics. An account with moderate keyword counts but dense audience layering burns through it faster than expected.
How many campaigns should one account have?
Technically up to 10,000. Practically, as many as can each receive enough conversions for a bid strategy to learn — roughly 30 conversions in 30 days as an orientation point. That puts the real ceiling two to three orders of magnitude below the technical one.
The interface is slow — is that a limit?
No, it is a consequence of volume. You are nowhere near a formal cap, but the web UI stops being practical somewhere in the low thousands of campaigns. Move to Editor and report-driven workflows.
What does the entity size error on upload mean?
Usually that a single operation or object is oversized, not that the account is full. Shrink the batch and look for an object carrying an abnormally long list of child items.
Do limits affect Performance Max?
Through assets, yes — asset groups draw on the account-wide asset allowance, and automatic generation accelerates consumption. Other limits touch PMax lightly because its structure is different.
Can I request a limit increase?
System limits are platform-wide and not negotiable, including the 2,500 accounts per week creation rate.
How do limits relate to the AI Max migration?
Directly, through the 25,000 DSA targets per account cap. Accounts built on dynamic search ads with large target sets should be reviewed before migrating: many targets lose their purpose, and pruning them belongs before the move, not after. Context in the DSA to AI Max migration.
Does removing entities free limit headroom immediately?
Yes, removed objects leave the count. But removal is irreversible and takes history with it, unlike pausing — so prune only what genuinely will not come back.
Is there a limit on conversion actions?
There is an explicit cap, but it is high enough that nobody reaches it. You will hit a different wall first: too many conversion actions dilute optimization because the system starts treating everything as a goal. That constraint is methodological, not technical.
