Your Agency Says Google Ads Is "Still Learning." Check It.
Your Google Ads numbers have gone flat, and the agency says the campaign is still learning and needs a few more weeks. Popular guides back the agency up: be patient, allow a week to a month, and leave targets and budgets alone because any tweak resets the clock. That advice turns a checkable claim into an open-ended wait, and you pay for every day of it. Google's own help pages changed their waiting guidance twice in four weeks and now disagree with each other, and this article shows you how to check the claim in your own account.
Google changed its own answer on how long to wait, twice in four weeks
PPC News Feed reported on August 27, 2026 that Google had cut the recommended wait after a Smart Bidding adjustment from "3 conversion cycles or 4 weeks" to "1-2 conversion cycles." The change hit the Target ROAS, Change Bid Strategy and Exploration help pages. Smart Bidding is Google's name for its automated bid strategies, the ones that set your bid in each auction from conversion data.
The second rewrite came weeks later. Anu Adegbola reported in Search Engine Land on September 22 that Google had rewritten its "Duration of the learning period" page. The old text said learning took "up to 3 weeks or 1-2 cycles." The new version frames it around conversion data. The UK and US English editions of that page now differ. The UK edition, read on October 2, 2026, says:
"It can take up to around 50 conversion events or 3 conversion cycles for the bid strategy to calibrate to the new objective, although it can be faster depending on the amount of conversion data present."
Google Ads Help, "Duration of the learning period," UK English edition, read October 2, 2026.
The US English edition of the same page, on the same day:
"It can take a few conversion cycles (1-2 typically) for the bid strategy to calibrate to the new objective, although it can be faster depending on the amount of conversion data present."
Google Ads Help, "Duration of the learning period," US English edition, read October 2, 2026.
A conversion cycle is the gap between someone clicking your ad and becoming a lead or a customer. Google uses it as the clock for learning. One edition tells a London owner to expect up to around 50 conversions or three cycles. The other tells a Chicago owner that one or two cycles is typical. Same page, same company, same day.
That gap hurts low-volume businesses most, and those are the accounts I know best. I have run Google Ads for 10 years, a lot of it for clinics and local-service businesses. One was a London ADHD clinic I booked solid for three straight months; they hired more specialists and outsourced the overflow. Businesses like that count success in booked appointments, and the monthly count is small enough to tally by hand. A "still learning" answer has to make sense at that volume or it isn't an answer.
Picture a physio clinic in Leeds that books, say, 25 new patients a month from search ads, with most people booking within a week of the click. The agency switches bid strategy and asks for patience. Under the UK wording, 50 conversions at that pace is two months of spend, though its three-cycle alternative reads closer to three weeks. Under the US wording, one or two weekly cycles is one or two weeks. The owner deserves to know which clock the agency is using. The table below shows how far apart the advice runs.
| Question | What Google's help pages say | What you'll be told |
|---|---|---|
| How long learning takes | UK English page: up to around 50 conversion events or 3 conversion cycles. US English page, same day: 1-2 conversion cycles typically. | HawkSEM: one week to one month. |
| Changing a target CPA or ROAS | Won't trigger a "learning" status and won't reset what Smart Bidding has learned. The bidder reacts immediately; give it 1-2 conversion cycles to hit the new target. | groas: a change of more than 20% resets learning. HawkSEM: target changes restart learning. |
| Changing the budget | Not named as a trigger on its own; the listed reasons are New strategy, Setting change and Composition change. | Navah Hopkins lists budget changes over 15% in a 7-day period among her triggers. HawkSEM: budget changes restart learning. |
Sources: Google Ads Help answers 13020501 (en-GB and en-US), 10970825 and 6268637, read October 2, 2026; HawkSEM (updated Nov 11, 2025); groas (Jun 16, 2026); Navah Hopkins, Search Engine Journal (Dec 11, 2025).
"Still learning" survives because a calendar can't be audited
Agencies did not invent the learning period. Navah Hopkins, a Microsoft employee writing a platform-agnostic column for Search Engine Journal in December 2025, described learning periods this way: "They're not monsters, and they're not imaginary." Her trigger list includes "Changing the budget more than 15% in a 7-day period" and "Changing TCPA/TROAS goals (especially if they are large changes)." That is careful practitioner caution from someone who works on a different ad platform, and it is reasonable.
The trouble starts when caution hardens into rules that Google's pages do not state. HawkSEM, in a guide updated November 11, 2025, tells readers learning takes one week to one month and that budget and target changes restart it. The groas blog went further in June 2026. It claimed that moving a target CPA or ROAS by more than 20% resets learning. Target CPA is the cost per lead you ask Google to aim for. Target ROAS is the revenue per ad dollar you ask it to aim for.
These rules persist because they suit everyone in the room. A waiting period is easy to promise and impossible to fail. If results improve in week four, the wait worked. If they don't, the campaign needs a little longer. The owner has no number to hold up against the explanation, only a date on a calendar that keeps moving.
Google's documentation adds a line that makes the phrase slippery: "Our algorithms continue to learn even when the bidding status no longer shows 'Learning'." Smart Bidding is always learning in the general sense. "It's still learning" is always technically true, and a statement that is always true explains no particular drop in leads.
The useful question is narrower. Did a specific change put this campaign into a Learning status? Has it seen enough conversions since then to recalibrate? The account answers these. The hover reason, the change history and the conversions counted since the change give the verdict.
Google's own pages contradict the target-reset rule
The page Google titles "How our bidding algorithms learn," read live on October 2, 2026, deals with target changes in a single sentence.
That sentence contradicts the 20% rule outright, and HawkSEM's line on targets along with it. Google's Target ROAS page adds the practical detail: "When you change targets, the bidder will react immediately but will need some time to hit the new target (give it 1-2 conversion cycles)." A bidder adjusting toward a new target keeps what it has learned. Your agency can fairly ask for one or two cycles after a target change. It can't fairly call that a learning reset, because Google's bidding page says the status won't show Learning for it.
When the status does say Learning, Google tells you why. Hover over it and the reason appears: New strategy, Setting change, or Composition change. Shopping campaigns can also show Ad group target change. Each reason describes a change someone made, and change history is where to look for it. A campaign on manual CPC, where you set the bids yourself, has no learning period at all. Google says the length of learning depends on the conversions obtained, the length of the conversion cycle and the type of bid strategy.
Search Engine Land's Anu Adegbola warned against treating the new number as a threshold: "The 50-conversion figure shouldn't be treated as a requirement that every campaign must hit before Smart Bidding works." She described the shift as a move from "How many weeks has it been?" toward "How much conversion data has the bidder actually seen?" An owner can answer that second question from a report.
Volume decides how fast any of this goes. Google requires at least 15 conversions in the past 30 days before Target ROAS can run on a Search campaign at all. Its UK learning page talks about up to around 50. A campaign that struggles to reach 15 a month is a long way from 50, and patience won't change that arithmetic.
Google's fix for thin data is in the same Target ROAS article: "Performance will generally improve if you run fewer, larger campaigns that get more conversions." Say your account splits 30 leads a month across six campaigns. Each of those campaigns learns from about five. If that sounds like your setup, read my guide on how many Google Ads campaigns you should actually have. For a low-volume clinic the answer is fewer than the agency built.
You can audit the claim yourself from the account
Log in yourself, or ask for read access if all you get is reports. These steps need no technical knowledge. You only need the account and a calendar for the dates.
Each step answers a question the agency should already have answered. The hover reason tells you what kind of change happened. Change history tells you when it happened and who did it. If the only recent change is a target adjustment, Google's bidding page says that does not produce a Learning status (Shopping ad group targets are the exception), and the agency owes you a different explanation. If no change appears near the date the status started, ask what triggered it.
Everything after that depends on the conversion count, and the count is only as good as your tracking. If the account counts page views, duplicate form fills or very short calls as conversions, the bidder is learning from noise. Confirm the count is real before you argue about learning. This breakdown of how broken conversion tracking makes a Google Ads account lie walks through the checks.
A brand-new campaign is a different case. Its problem is how much budget it burns before any data exists, which I cover in why a big Google Ads launch usually burns your cash. This audit is for later, when an established campaign changed something and the explanation for weak results became "give it time."
Now compare what you found with both English editions. Say a campaign has 40 conversions and two cycles since a strategy switch. It has passed the US guidance and is close to the UK figure, and "still learning" no longer explains much. A campaign with six conversions and one cycle has a fair claim to more time. The next question for that campaign is whether it will ever get enough conversions.
The conversion count since the change decides it
Measure on a cycle clock. Write down the date of the change that triggered learning. Check the campaign after one cycle, after two, at three, and again once conversions since that date approach 50. Those checkpoints follow Google's wording in each edition. If performance hasn't moved by the last of them, learning stops being a believable explanation, and the strategy itself is the question.
Watch for metrics that look like progress. Clicks and impressions can climb while leads stay flat. A falling cost per click feels like a win until you divide spend by booked appointments. The Learning badge going away doesn't prove anything either. Google says its algorithms keep learning after the status clears, and the badge doesn't measure customers.
Rule out the other causes too. From August 17, 2026, Google changed how it bids for budget-limited Target CPA and Target ROAS campaigns and warned of "temporary performance and traffic fluctuations." A campaign held back by its budget could have wobbled in late August without ever showing a Learning status. That one was a Google-side change. Your agency should be able to tell you whether your campaigns were budget-limited at the time.
Take these questions into your next call with the agency. A good agency answers them in a minute with the account open.
The last question matters most for clinics and local services. If the campaign cannot reach that kind of conversion count in any reasonable window, waiting is a cost with no end date. The fix is structural: fewer campaigns, pooled data, or a bid strategy that matches the volume you actually have.
Frequently Asked Questions
How long should the Google Ads learning phase take?
It depends on your conversion volume and how long customers take to convert, and Google's own pages currently disagree. On October 2, 2026, the US English help page said one or two conversion cycles is typical. The UK edition said up to around 50 conversion events or three cycles. A conversion cycle is the time from ad click to conversion. Judge it by the conversions since the change that triggered learning, using your change history.
Does changing my budget or target CPA reset Google Ads learning?
Google's bidding help page says a target change does not. It states: "Changing a target won't trigger a 'learning' status, and won't reset anything Smart Bidding has already learned about your account." Google's learning page does not name budget changes as a trigger on their own. Practitioners still urge caution with big swings; Navah Hopkins lists a budget change over 15% in a 7-day period among her learning-period triggers.
What if my Google Ads campaign never gets out of learning?
Check the volume first. A campaign that can't reach 15 conversions in 30 days can't even use Target ROAS on Search, and Google's UK learning page puts calibration at up to around 50. Google's own advice is that performance generally improves with fewer, larger campaigns that get more conversions. Ask your agency to consolidate or to switch to a strategy that fits the conversions you actually get.
Before your next agency call, open change history and note the date, the author and the conversions since. If the numbers support patience, give it with a clear checkpoint. If they don't, you have Google's own pages to point to. Google can't agree with itself on how long learning takes. Ask why your agency sounds certain.
