Ad Refresh Best Practices: More Impressions Without Killing Your CPM

By K. A. M. Rashedul Mazid — Publisher Growth · 9 min · August 2026

Ad refresh is the most misunderstood lever in publisher monetisation. Done well, it converts long reading sessions into extra revenue that costs you nothing. Done badly, it inflates impression counts, tanks your viewability, annoys buyers, and eventually gets your inventory quietly deprioritised. The difference is entirely in the rules you set.

Key takeaways

Refresh on user engagement and viewability, never on a blind timer.
30 seconds in-view is the industry-safe minimum interval; below 20 seconds you're inviting buyer complaints.
Cap refreshes per slot (typically 3–6) so a single reader doesn't generate twenty impressions on one unit.
Exclude sponsorship, programmatic guaranteed and any creative with a view-through commitment.
Measure the outcome in RPM, not impressions — impression growth with falling CPM is a net loss.

Key numbers

20–50% — Typical impression uplift from correctly configured refresh on long-session sites (Google Ad Manager publisher guidance)
30s — Minimum in-view interval most demand partners consider acceptable (Google Ad Manager refresh policy)
50% / 1s — MRC standard for a viewable display impression (Media Rating Council viewability standard)
2s — MRC in-view requirement for video impressions (Media Rating Council)

What Refresh Is (and What Buyers Think It Is)

Refresh replaces the creative in an existing slot with a new auction result. From your side it's extra inventory. From the buyer's side it's a question of whether the second impression was actually seen by a human who was still paying attention.

That's the whole tension. Buyers aren't opposed to refresh — they're opposed to refresh that manufactures impressions nobody looked at. Every rule below exists to prove your refreshed impressions are real.

The Rules That Keep Refresh Profitable

Rule one: only count in-view time. If the slot is off-screen, the timer pauses. A reader who scrolled past your top banner two minutes ago should not be generating banner impressions.

Rule two: 30 seconds minimum. Google Ad Manager's own guidance treats 30 seconds of in-view time as the safe threshold, and most SSPs align with it. Some publishers push to 20 for sticky footers; below that you're gambling.

Rule three: cap it. Three to six refreshes per slot per pageview keeps the ratio believable. Uncapped refresh on a two-hour live blog produces numbers that get flagged.

Rule four: require actual engagement. Tie the timer to scroll, tab focus, or interaction. A backgrounded tab should not refresh anything — that's the single behaviour most likely to get you labelled as invalid traffic. The wider fraud-hygiene picture is in ad fraud protection and yield .

Rule five: exclude direct. Any programmatic guaranteed, sponsorship or fixed-price line item should sit outside the refresh logic entirely. Refreshing a sponsor's creative is the fastest way to lose a renewal.

Measuring Whether It Worked

Impressions will go up. That's not the test. The test is session RPM — revenue per thousand sessions — before and after, held over at least two weeks.

You'll usually see CPM drop somewhat, because later impressions in a session are worth less to buyers. That's fine, as long as total revenue rises. If CPM falls 40% while impressions rise 30%, you've made yourself poorer and your inventory less desirable at the same time.

Watch viewability alongside it. If it slides below roughly 60%, tighten the interval or reduce the cap before demand partners notice. The mechanics of that metric are covered in the viewability and vCPM guide .

How to Implement It Without Breaking Things

Most publishers implement refresh through their Prebid wrapper or via a Google Publisher Tag event listener that watches impressionViewable, starts a timer, checks the slot is still in view, then calls refresh on that specific slot.

Two implementation details save a lot of pain: refresh one slot at a time rather than the whole page, and pass a refresh counter as a key-value so you can segment reporting later. Without that counter you'll never be able to prove which impression tier is performing.

Roll it out on one template first — usually long-form articles — and leave the rest of the site as a control group for a fortnight.

Glossary

Ad refresh: Requesting a new ad for a slot the reader has already seen, without a page reload.
In-view time: How long at least 50% of the ad has been visible on screen — the clock that should drive refresh.
Session depth: How long an average reader stays. The longer the session, the more refresh is worth.
Refresh cap: The maximum number of refreshes allowed per slot per pageview.
vCPM: Cost per thousand viewable impressions — the metric buyers increasingly pay on.

Frequently asked questions

Is ad refresh against Google's policy?

No. Google Ad Manager permits it, provided the refresh is tied to genuine in-view time or user action and isn't used to manufacture impressions. Blind timers on off-screen slots are what get publishers into trouble.

What is the safest refresh interval?

30 seconds of in-view time. It's the threshold Google's guidance uses and the one most SSPs are comfortable with.

How many times should one slot refresh?

Three to six per pageview for most sites. Live blogs and forums can justify the upper end; standard articles rarely need more than three.

Will refresh lower my CPM?

Usually a little. Later impressions in a session attract lower bids. That's acceptable as long as total session revenue rises — measure RPM, not CPM.

Should I refresh sticky footer ads?

Sticky units are the best refresh candidates because they stay in view by design. They're also the ones where a short interval is most tempting and most risky, so keep the cap tight.

Does refresh hurt viewability scores?

It can, if you refresh slots that drift out of view. Gating on in-view time protects the metric; blind timers destroy it.

Should refresh run in background tabs?

Never. Use the Page Visibility API to pause everything when the tab loses focus. Background refresh is the fastest route to an invalid-traffic problem.

Can I refresh video ads?

Not in the same way. Video is governed by pod rules and completion metrics; use sequential ad pods rather than refresh logic.

Do direct sponsors need to be excluded?

Yes, always. Guaranteed and sponsorship line items should be flagged non-refreshable in your ad server so a sponsor's creative is never rotated out early.

How long should I test before deciding?

Fourteen days minimum, with one template as a control. Weekly seasonality and buyer pacing make shorter tests unreliable.

What warning signs mean I've gone too far?

Viewability under 60%, rising invalid-traffic deductions, SSPs reducing bid volume, or a CPM drop steeper than your impression gain.

Is refresh worth it for a small site?

Only if sessions are long. On a site averaging under a minute per visit, refresh adds noise and risk with very little revenue to show for it.