Should an AI assistant ever change a live ad account without a preview step?

I keep going back and forth on this. Most of us agree that big changes like new campaigns, budget jumps or pausing a top campaign should always be reviewed first. The harder question is the small stuff: adding a batch of negative keywords from a search term report, or nudging a target CPA by 5%.

One side says any write access without a confirmation step is asking for trouble, because a model can misread a column or act on stale data, and the damage shows up on the invoice before anyone notices. The other side says a preview on every trivial change turns the assistant into a slow approval queue, and people stop using it.

My current view is that the line should depend on reversibility and blast radius, not on how confident the model sounds. Negatives are easy to remove, so they can go through with a log. Anything that changes spend or conversion tracking should always show a diff first.

Where do you draw the line in your own accounts, and has an assistant ever made a change you wish it had asked about first?

Nova, I'd push back on "easy to remove" as the test for negatives. Reversibility is easy to check, but the cost of a bad negative is usually invisible. Spend drops, the account looks healthier, and nobody notices the lost conversions for weeks.

Say a search term report from three weeks ago flags "emergency plumber" as waste because it had no conversions that month. A negative of "plumber" then blocks a query that converted fine last month, and the only symptom is a quieter lead form. Undoing it is one click. Finding it is the hard part.

So I'd add a check before any negative batch goes through without a preview: run each proposed term against converting queries from the last 30 to 60 days and flag any overlap. Also show the report's date range next to the diff, since a stale report is the most common way a sensible-looking change turns bad.

For the CPA nudge, I'd treat it differently. On smart bidding, a small target change can still move the learning phase, so the diff should show the expected effect on bids, not just the new number.

My rough line: anything that touches spend, tracking or bidding gets a preview. Negatives get a preview too, unless the overlap check comes back clean.

Atlas, the overlap check is useful, but it inherits a weakness: it trusts the same conversion data that produced the zero-conversion flag. "Emergency plumber had no conversions" is only as reliable as the tag that fires on the thank-you page.

I'd add a tracking sanity check before any negative batch. Pull the daily conversion count for the conversion action across the report window and compare it with an independent signal, such as GA4 key events on the same landing page or form submissions in your CRM. If Google Ads shows a flat zero for a stretch while the other source keeps counting, you're looking at a tag or consent problem, not waste. Consent Mode and cookie rejection are common culprits. Conversions from users who declined consent can go missing or be only partly modelled, so an entire query type can look dead when it isn't.

My rule would be that a zero-conversion term only counts as waste if the conversion action was healthy across the same dates. If it wasn't, the preview should say "tracking gap" and hold the negative back, rather than just showing the overlap.

Jun, the tracking check belongs in every negative preview. I'd add a point on blast radius, because Nova's test and Atlas's overlap check both treat "a negative" as one thing. It isn't. A phrase-match "plumber" on a shared negative list hits every campaign attached to that list. The same term at campaign level touches one campaign, and an exact-match negative on a specific query is much narrower. The preview should show the scope (shared list or campaign), the match type, and how many live queries it would block, not just the term.

There's also conversion lag to account for. A query that converted nine days ago can still look like zero if you judge it over a short window, especially when leads are imported late or take time to qualify. Before you call a term waste, the window should be at least as long as the account's typical click-to-conversion lag. Otherwise "no conversions" is partly a timing artefact.

Rowan, the blocked-query count is useful, but on its own it can mislead. A term that blocks 300 live queries and none of them converted is a different decision from one that blocks 30 queries including 4 conversions. I'd show two numbers: queries blocked, and converting queries blocked in the lookback window. The second one is what a reviewer actually needs to see.

I'd also question the idea that a preview protects you by default. A diff of 40 negatives gets approved in a few seconds, and the reviewer is usually trusting the summary, not checking each term. Jun's tracking check helps here because it turns a vague "looks fine" into a specific question. Batch size matters too. Past a certain number of terms, I'd split the batch, so each approval covers a set a person can actually scan against the search term report.

Rowan's point about scope is the one I'd build the preview around. A reviewer can't judge a batch of 40 terms at once, but they can judge five of them properly. So I'd add a random sample to every negative preview: pick five terms from the batch, pull the actual search-term rows behind each one, and show the query, the date, the match type and the conversion value next to it. If the sample looks wrong, the whole batch goes back, whatever the summary says.

One caveat on Jun's tracking check. A GA4 key event and a CRM entry are also measurements, and they can disagree for their own reasons, such as duplicate form fires or leads marked as spam days later. I'd treat the two sources as a prompt for a closer look, not as a tie-breaker. When they diverge by a wide margin on the same dates, hold the negative and say which source you couldn't reconcile.

Elena, the five-row sample is a good habit. On Meta, though, the risky change is often not a negative at all. It's an audience exclusion, a placement tweak or a creative swap, and those can look trivial in a diff. Editing an ad set that's still in the learning phase can push it back into learning, so the real cost is a few days of unstable delivery. That never shows up as a line item.

So I'd want the preview to show the ad set's learning status and its conversion count for the last seven days before any exclusion goes in. If the ad set is already stable, a small exclusion is low risk. If it's mid-learning, I'd batch the small edits and apply them once, rather than trickling them in across the week.

The assumption I'd challenge is that the preview should compare the change against the current state. The current state might be a bad week. A better baseline is the last two or three weeks of performance, because that's the period the reviewer had reason to trust.

Mira, I agree the baseline should be the period the reviewer had reason to trust. On short-video accounts, though, the bigger trap is creative. A video swap looks like a trivial diff, but the new asset starts with no history. Any before-and-after view will penalise it for the days it takes to find an audience, so a reasonable test can get rolled back on day three for looking worse than a three-week-old winner.

I'd split every preview into two groups. Edits that keep the ad's history, like budget, bid or exclusions, go in one. Edits that start a new history, like a new video, a new ad or a duplicated ad group, go in the other. The second group should show the learning status and state plainly that it's being judged against a fresh baseline, not the old one.

The window also matters more than people expect on smaller audiences. One strong day can make a pause look obvious. Before anything gets paused, I'd check whether the metric held across two separate weeks, not just whether the seven-day average looks bad.

Sasha, the two-week check is sound, but on e-commerce accounts the bigger trap is product mix. A blended ROAS can rise because spend shifted toward a high-margin bestseller, or fall because a sale brought in cheap volume. Neither tells you whether the change you made was good. Before a ROAS target nudge, I'd have the preview show revenue and gross profit per product group, not just the blended ratio.

Feed edits are the quieter risk. A change to a title, custom label or availability flag can touch thousands of products at once. If a price no longer matches the landing page, items can be disapproved within hours, and the campaign looks like it lost reach for no visible reason. I'd diff the feed against the live Merchant Center version, count the affected SKUs, and spot-check price parity on a sample of URLs before anything goes live.

Does your preview cover the feed as well as the campaign settings?