Which Google Ads changes put a bid strategy into “Learning”—and does a new ad count?
Google lists a few Learning triggers, and a new Search ad isn't one. See what each Google source says, what resets learning, and how long to wait before judging.
Google names a short list of reasons for the Learning label, and in Search a new ad isn't on it. But the label isn't the learning, Google's pages disagree about what resets it, and "how long should I wait?" has different answers depending on what you're waiting for. Here's how to check a change against Google's own sources before you make it.
Everything attributed to Google here is as of October 14, 2026. Each source is linked where it's used and listed, with that date, at the end.
Google's documentation names only a few reasons a bid strategy shows "Learning": the strategy is new or was reactivated, one of its settings changed, or campaigns, ad groups or keywords were added to it or removed from it. Google's API status codes add budget changes and changes to conversion types or conversion settings. Adding an ad appears on none of these lists (About bid strategy statuses; Google Ads API status codes).
That settles the label question for Search. It doesn't settle the question you probably have to act on: should you make a change now or wait one out, and if you wait, for how long? That needs two more things: what the bidder does when there's no label, and which of Google's several waiting periods applies to you.
A label, and the learning underneath it
Two different things get called "learning." Most confusion about what "resets" a campaign comes from treating them as one.
The first is a status. Google says that after you change your bid strategy, "there may be minor performance fluctuations as Google Ads optimizes your bids," and "to indicate this, a 'Learning' status may be shown." Hovering over the status shows which reason applies (About bid strategy statuses).
The second is the bidder's continuous adjustment. Google says Smart Bidding "is always learning," refines its models with "conversion rate data at more granular levels (e.g. for a search query mapped to specific ad copy or landing pages)," and names "adding new keywords, testing new ad copy or updating landing pages" as common internal causes of fluctuation (How our bidding algorithms learn). Its learning-duration page adds that the algorithms "continue to learn even when the bidding status no longer shows 'Learning'" (Duration of the learning period).
Together these mean the label tells you Google recognized one of its listed change types. It doesn't measure how much a change matters, and when the label is absent, learning still goes on.
Much of Google's timing guidance, though not all of it, is given in conversion cycles: "the typical time it takes for a click to convert" (How to make target adjustments with Search Smart Bidding). The cycle includes the time imported conversions take to be reported (How our bidding algorithms learn). If most of your conversions are reported within seven days of the click, your conversion cycle is about seven days.
What Google names, change by change
Find the row for the change you're planning. The middle column says whether that change is one of Google's listed reasons for the Learning status. There are two such lists:
- The Help status list: the reasons that Google's two Help pages about the status say the hover text can show.
- The API status codes: the more detailed values the Google Ads API reports.
Other Google pages appear in the right-hand column.
Google's sources on common changes, as of Oct. 14, 2026
| Change | A listed Learning reason? | What else Google says |
|---|---|---|
| Create a bid strategy, or reactivate one | Yes. Help status list: "New strategy: The bid strategy was recently created or reactivated." API: LEARNING_NEW. |
For resuming a paused campaign, see the note below the table. |
| Change a bid strategy setting | Yes. Help status list: "Setting change." API: LEARNING_SETTING_CHANGE. |
Neither list says which settings count. For CPA and ROAS targets, see that row. |
| Add or remove campaigns, ad groups or keywords | Yes. Help status list: "Composition change." API: LEARNING_COMPOSITION_CHANGE, a "change in number of campaigns, ad groups or keywords attached to it." |
— |
| Change the budget | API code only: LEARNING_BUDGET_CHANGE. Not on the Help status list. |
A separate Help page, on conversion goals, says frequent budget changes during a "7-14 day learning phase" "reset the learning window." That page doesn't mention the label. |
| Change conversion goals or actions | API codes only: LEARNING_CONVERSION_TYPE_CHANGE and LEARNING_CONVERSION_SETTING_CHANGE. |
Smart Bidding takes "1-2 conversion cycles in most cases" to learn after any change to goals or actions. |
| Change a campaign or portfolio CPA or ROAS target | No, in Google Ads: changing a target "won't trigger a 'learning' status." | Whether it resets learning is disputed (see "Target changes"). Search Ads 360 says significant target changes can trigger learning in strategies it manages. |
| Change an ad group target in a Shopping campaign | Sometimes: "In some cases, ad group target changes may trigger a 'Learning' status on Shopping campaigns." Google gives no condition. | Listed only on the learning-duration page, not the status page. |
| Add or change an ad | No, in Search. Ads aren't on any list. | Display and Demand Gen are different (see "The new-ad case"). |
| Turn on Smart Bidding Exploration (an opt-in setting that accepts a lower Target ROAS to reach new searches) | No: "There is no learning period." | A ramp-up "may be 1-2 weeks" if a lot of new valuable traffic is found. |
| Bid with Manual CPC | Not applicable: the learning period "isn't applicable to Manual CPC." | — |
Table sources:
- About bid strategy statuses
- Duration of the learning period
- API status codes and proto
- About conversion goals
- Changing conversion goals and actions
- How our bidding algorithms learn
- Search Ads 360 learning page
- Smart Bidding Exploration FAQ
Treat these lists as Google's description, not a specification, because Google's own counts don't match. The status page promises "4 reasons" and lists three. The learning-duration page promises three reasons and lists four, in both its US and British English editions: the status page's three, plus the Shopping ad group target change in the table. The API also reports budget and conversion changes that the Help status list leaves out. The hover text on your own strategy is better evidence than any of these lists.
Resuming a paused campaign. Google's list treats a strategy that was "recently created or reactivated" as new, and the same page names paused campaigns as one reason a strategy shows Inactive. Google doesn't say in so many words that resuming a paused campaign reactivates its strategy, or how long the pause has to be. Reading the two statements together, expect that a resumed campaign may show Learning with the "New strategy" reason, and confirm it in the hover text. If a resumed campaign isn't delivering at all, the answer on restarting after a 15-day pause covers what to check first.
The new-ad case
Take a constructed example: a Search campaign on Target CPA, and you want to test a second ad.
If you add the ad to an existing ad group, you've changed no strategy setting, no budget, no goal, and no count of campaigns, ad groups or keywords. Nothing on Google's lists predicts a Learning label. That doesn't mean the bidder ignores the ad. Google lists "ad characteristics" among Smart Bidding's signals: it can adjust bids "based on which version of an ad will be shown" (About Smart Bidding). Google also names new ad copy as a common cause of fluctuation, and advises letting "at least one conversion cycle" pass before you evaluate performance (How our bidding algorithms learn).
Now suppose you put the new ad in its own new ad group to keep the test tidy. The lists cover ad groups "added to … the bid strategy" (Help) or "attached to it" (API). In a campaign with its own standard strategy, a new ad group is attached to that strategy. That is our reading rather than a sentence Google writes, but it's a direct one. On that reading, the same test, packaged differently, now matches a listed reason: Google's list responds to how a change is packaged, not to the ad itself.
Outside Search, Google is more cautious about creative, so the answer above doesn't carry over:
- Display. The Display Smart Bidding guide lists "new creatives" among the "significant campaign changes" that "can disrupt performance because the algorithm will need to learn about and adjust to the new settings." It advises making significant changes all at once to reduce learning time, and says you can "check the status column to see if the algorithm is still learning." It doesn't say whether the label appears after a creative change.
- Demand Gen. The Demand Gen performance guide says "changing settings forces the campaign to do some or all learning over again." A creative change also sends the ad back through policy review, which takes "around 24-48 hours."
So the no-label answer is documented for Search. In Display and Demand Gen, assume the system relearns from new creative, and check the status column rather than assuming either way about the label.
Target changes: two answers about resets
On the label, Google Ads speaks with one voice: "Changing a target won't trigger a 'learning' status" (How our bidding algorithms learn). The only Google Ads exception is the Shopping ad group target line in the table. Search Ads 360 differs for strategies it manages: "significant changes" to a target can trigger a learning status there (Search Ads 360 learning page).
Where Google disagrees with itself is on whether a target change resets learning:
- No reset. The algorithms guide says a target change "won't reset anything Smart Bidding has already learned about your account." It says Smart Bidding "reacts immediately to any target change," and that "in some rare cases" you may see volatility "for a conversion cycle or two."
- Reset. Google's conversion-goals page says Smart Bidding needs "a standard 7-14 day learning phase," and that "frequent manual changes to budgets, targets (CPA or ROAS), or conversion goals during this period … will reset the learning window."
No page reconciles the two.
In practice, the guides agree on when to judge a target change but divide on how often to make one. Google's Search guide to target adjustments says the bidder starts optimizing "within minutes" but may need "1-2 conversion cycles" to hit the new target "because of the delay in conversions." It asks you, "unless there is a true business need," not to make multiple ROAS target changes within a single conversion cycle, for two reasons: changing "before 100% of conversions have been reported makes it impossible to follow best practices for performance assessment," and changing mid-cycle "impedes performance." Yet the same page says Smart Bidding "can perform well with rapid target changes," and the algorithms guide says you "should feel comfortable changing CPA and ROAS targets as frequently as you would like."
So a Google Ads target change brings no label and an immediate reaction. You can't judge the results until the conversions have arrived. Google permits frequent target changes, but if you want to read a ROAS change in Search, its guide gives documented reasons not to stack another change inside that window.
How long to wait, and by what measure
The label is supposed to help with exactly this decision. Google gives several answers because it is answering different questions.
What Google says about waiting, as of Oct. 14, 2026
| What you're waiting for | Google's measure | Source |
|---|---|---|
| The bidder to hit a new target (Search) | 1–2 conversion cycles, "because of the delay in conversions" | Target adjustments guide |
| Calibration after a bid strategy change | US English edition: "a few conversion cycles (1-2 typically)." British English edition: "up to around 50 conversion events or 3 conversion cycles." | Learning-duration page: US, UK |
| Learning after a conversion goal or action change | "1-2 conversion cycles in most cases" | Changing conversion goals and actions |
| The learning phase, as the conversion-goals page states it | "a standard 7-14 day learning phase" | About conversion goals |
| A fair evaluation of Smart Bidding results | Periods "that have at least 30 conversions, such as a month or longer (50 conversions for Target ROAS)" | About Smart Bidding |
| Judging a Demand Gen change | "at least 30-50 conversions" | Demand Gen guide |
| Search Ads 360 Target CPA or ROAS strategies | "2-3 conversion delay cycles" | Search Ads 360 learning page |
The two English editions of the same learning-duration article give different calibration figures, and Google doesn't say why. If you've met "50 conversion events or 3 conversion cycles" elsewhere, the British English edition is where Google says it. Neither edition says what clears the label.
The measures come in two kinds:
- Cycle measures tell you when the conversions you'll judge have been reported.
- Count measures tell you when there are enough of them to judge.
Our synthesis, not a Google rule: a fair judgment needs both, so whichever takes longer for your campaign sets the wait.
The cycle clock. A fixed calendar rule fits only some accounts. Here is Google's one-to-two-cycle range applied to three constructed campaigns. These are examples, not predictions:
| Constructed campaign: most conversions reported within… | 1–2 conversion cycles is… | Compared with "7–14 days" |
|---|---|---|
| A: 2 days | 2–4 days | Shorter |
| B: 7 days | 7–14 days | The same |
| C: 21 days (for example, imported offline sales) | 21–42 days | Up to three times longer |
For campaign C, a two-week rule would have you judging results before one full conversion cycle has passed.
The count clock. Give the same constructed campaigns a conversion volume:
| Constructed campaign | Conversions per month | A period with 30 conversions takes about… |
|---|---|---|
| A (2-day cycle) | 15 | 2 months |
| B (7-day cycle) | 60 | 2 weeks of clicks, plus up to 7 more days for them to be reported: about 3 weeks |
| C (21-day cycle) | 90 | 10 days of clicks, plus up to 21 more days for them to be reported: about a month |
Here's how the two clocks compare for each campaign:
- A: the count governs. Two months, against a cycle measure of two to four days.
- C: the wait is about a month to six weeks. The count clock, mostly reporting lag, sets the minimum at about a month (31 days). At the top of Google's one-to-two-cycle range, the cycle clock runs longer, to 42 days.
- B: the count clock sets the wait at about three weeks (22 days), longer than even two conversion cycles (14 days). Google's "such as a month or longer" is its example of a period with 30 conversions, not a separate clock.
For Target ROAS, Google's figure is 50 conversions, which lengthens every count.
How often to change. Google's pacing advice differs by product. Read as a set (again, our reading; Google doesn't forbid frequent changes), it follows one logic: don't change faster than you can read the last change.
- Search: unless there's a true business need, avoid more than one ROAS target change within a conversion cycle. For CPA and ROAS alike, Google also says you can change targets "as frequently as you would like."
- Display: change bids by 20% and wait a week between changes. "Small, infrequent" Target CPA or ROAS changes "aren't likely to cause a noticeable problem," and an ad group with "more than 100 conversions per day" can change more often. That fits the same logic: with that volume, its results become readable sooner.
- Demand Gen: flex bids "by 5-10% (no more than 15%) per week," and "refrain from making any changes whatsoever during the initial learning period."
- Search Ads 360: adjust targets gradually, waiting "at least 2-3 conversion delay cycles for each adjustment."
Each interval is that product's waiting measure, applied to its own last change. Sources: the target adjustments guide, the Display guide, the Demand Gen guide and the Search Ads 360 learning page.
When the label clears. Google's instruction for the period is modest: "Continue using your account as usual," but "you may not want to measure performance until the learning period is over" (About bid strategy statuses). A cleared label doesn't tell you either clock has run. Set your wait by the measures, and write down before the change the date or conversion count at which you'll judge it.
If you're looking at a Learning label now
Find the reason. In Google Ads, the Bid strategies page under Tools has a Status column for portfolio strategies, and hovering over the status shows the reason. Through the API, campaign.bidding_strategy_system_status reports status for both standard and portfolio strategies. A value of MULTIPLE_LEARNING means more than one reason applied during the period (About bid strategy statuses; API bid strategy status guide).
Then match the reason to what actually changed. If the label says composition change and you only touched ads, find out who or what added or removed keywords or ad groups. It might have been a colleague, an automated rule, a script or an auto-applied recommendation.
Then decide which clock you're on. Use the waiting table above, not the label, to decide when to judge.
The label also isn't a diagnosis. A resumed campaign that shows Learning and gets no impressions needs the serving checks in the 15-day pause answer. If you're bringing back old campaigns after a restructure, the answer on restoring merged campaigns covers what reactivation can and can't recover.
If you were hoping to trigger Learning
Sometimes the instinct runs the other way: after a bad stretch, push the strategy into Learning so it starts fresh. No Google source documents a benefit from the label itself. The changes that produce it bring their own relearning cost:
- Google says Smart Bidding needs "1-2 conversion cycles in most cases" after a change to conversion goals or actions.
- Its Demand Gen guide calls changing the conversion goal on a live campaign "the biggest reset to learnings," and says settings changes force "some or all learning over again."
- The conversion-goals page says frequent changes reset the learning window.
These resets are a cost of the change, not a cleanup. A target change won't clear a bad stretch either, since Google says it doesn't discard what was learned. It changes the instruction, not the history.
If the bad stretch came from broken conversion tracking, that's a different problem. The answer on restoring merged campaigns explains when data exclusions apply. If the conversion goal itself is wrong, Google's guide to changing conversion goals and actions sets out transition workflows, and the answer on low-volume B2B campaigns covers choosing the lead outcome first.
Where the sources go quiet
Several common edits sit outside every list, and Google says neither that they produce the label nor that they can't: negative keywords, location or audience targeting, ad schedules, bid adjustments, assets and landing pages. It also doesn't say:
- whether pausing (rather than removing) keywords or ad groups counts as a composition change;
- which strategy settings count as a "setting change," or how large a change has to be;
- why its two English editions give different calibration figures.
For those changes, the evidence is your own hover text, read against your change history.
Sources
All as of October 14, 2026.
- Google Ads Help, About bid strategy statuses
- Google Ads Help, Duration of the learning period for campaigns and what affects it (US English edition) and its British English edition
- Google Ads Help, How our bidding algorithms learn
- Google Ads Help, How to make target adjustments with Search Smart Bidding
- Google Ads Help, About Smart Bidding
- Google Ads Help, About conversion goals
- Google Ads Help, Changing conversion goals and actions used for Smart Bidding
- Google Ads Help, Manage your Smart Bidding strategy for Display campaigns
- Google Ads Help, Demand Gen campaign performance guide
- Google Ads Help, Frequently asked questions about Smart Bidding Exploration
- Search Ads 360 Help, About bid strategy learning in Search Ads 360
- Google Ads API v26, BiddingStrategySystemStatus, its proto definition and the bid strategy status guide