To explain Performance Max asset results to a client, export the asset report, give the AI only the columns it needs (asset name, impressions, clicks, cost, conversions and any performance label shown), and ask for a plain-English summary that separates what Google measured from what you recommend. Then check every claim against Google's own caveats before it reaches the client.
The caveat that matters most: asset numbers and ratings are comparisons, not grades. Google counts asset metrics once per asset shown, calls asset-level ratios directional, and describes its performance labels (Best, Good, Low) as relative rankings against other assets of the same type. A "Low" asset can still be producing conversions, and a "Best" one can be producing none that you care about. An AI summary that treats either as pass or fail will mislead your client confidently.
What the asset report actually shows, and what it hides
The asset report lists every asset used in a Performance Max campaign (headlines, descriptions, images, logos, videos) so you can compare them at the level of a single asset, an asset group, a campaign or several campaigns together. Google says the point is to decide which assets to rotate, remove or improve over time. It is a steering aid, not an audit trail of who saw what.
Three limits from Google's own asset-level metrics page should shape every client explanation. First, asset-level metrics are attributed per instance of the asset served within an ad. If one impression shows three assets, each of the three registers an impression, so the asset rows will not add up to the asset group or campaign totals. Second, ratio metrics such as click-through rate, cost per click, cost per acquisition and return on ad spend are directional only, because each asset is always served alongside others and the combination drives the result. Third, Google's own best practice is to judge performance at asset group or campaign level, not asset by asset.
The asset group report adds an "Added by" column showing whether an asset was uploaded by you or created or enhanced with Google AI, and a network segmentation (with search partners) that lets you see how assets performed across channels such as YouTube, Display, Search, Discover, Gmail and Maps. Both are useful for client conversations. "Added by" settles the question of whether the headline that did well was yours or Google's. The channel split explains why an image that looked weak in the total might be doing its job on one surface.
| What the client sees | What it means | What it does not mean |
|---|---|---|
| Label "Low" on an asset | Ranked below other assets of the same type in this campaign | That it loses money, or that it should be deleted today |
| Label "Best" on an asset | Ranked above its siblings | That it caused the conversions in the campaign total |
| "Learning" or "Pending" | Learning: impressions have started but there is not yet enough data to rank it. Pending: no performance data yet, perhaps still under review | That something is broken |
| Asset impressions larger than the campaign's | Counted once per asset shown in an ad | That the report is wrong |
| Ad Strength "Excellent" on the asset group | The group has every asset type, including optional ones such as sitelinks and extra logo and image formats | A guarantee of better results; it measures how complete the asset set is |
Check what your account actually shows before you build a template around it. Google stopped returning performance labels for Search and Display campaigns in its January 2026 API release, its current Performance Max help pages describe asset-level metrics rather than labels, and the rating wording differs between campaign types. If a current export has a label column, put its real values into your prompt; if it has none, build the explanation on the metrics alone.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowExport the data without handing over more than you must
Select Assets from the Campaigns menu, choose Assets again from the Assets drop-down, pick the Performance Max campaign from the Campaigns drop-down at the top, and download the table. For asset-group-level numbers, go to Asset groups, pick the campaign, switch the view from Summary to Table, and add columns through the Columns button. Google's recommended metrics for that report are conversions, conversion value, conversion rate and cost per conversion, which make a sensible minimum set.
Before the file goes near an AI tool, strip it down. Keep asset type, asset text or file name, the label if your export has one, impressions, clicks, cost, conversions and conversion value, and the date range. Delete any column that carries customer information. If a headline contains a promo code, a staff name or a customer quote, replace it with a neutral tag such as "headline 4". Ten minutes of tidying means the prompt works the same way for every client, and nothing sensitive leaves your account.
Add one line to the top of the file that states the date range and the date of the last asset change. Without it, the AI cannot tell a headline that has run for ten weeks from one added on Monday, and it will compare them as if they were equal.
The main example: a gift shop's autumn campaign
Say your client is a gift shop that sells hampers and cards online, with one Performance Max campaign, a monthly budget of $3,000 and three asset groups: Hampers, Cards and Corporate gifts. You are writing the November report. In the Hampers group the export shows the following (illustrative figures, from an account whose export includes a label column).
| Asset | Type | Label | Impressions | Cost | Conversions |
|---|---|---|---|---|---|
| Headline 2: "Hampers delivered in 2 days" | Headline | Best | 41,800 | $310 | 9 |
| Headline 5: "Gifts for every budget" | Headline | Low | 38,200 | $290 | 1 |
| Headline 7: "Order by Friday" | Headline | Learning | 1,900 | $14 | 0 |
| Image 3: hamper on a table | Image | Best | 52,000 | $405 | 11 |
| Image 6: gift wrap close-up | Image | Low | 9,300 | $61 | 0 |
| Video 1: 15-second unboxing | Video | Good | 27,500 | $188 | 4 |
The prompt below is the one I would give an assistant. Notice that it fixes the reader (a shop owner with no ad-account knowledge), orders the job (facts first, then interpretation, then recommendations), and tells the AI what it may not claim.
You are drafting the Performance Max asset section of a monthly client report.
Reader: the owner of a gift shop. No ad-account knowledge. Reads on a phone.
Data below is an export of one asset group for 1-30 November. Asset labels are
Google's relative rankings within this campaign, not pass/fail grades.
Asset-level numbers are counted per asset shown, so they do not add up to the
campaign total. Ratios per asset are directional only.
Write:
1. Three sentences: what happened, in plain English, using only the figures given.
2. A table: asset, what Google's label says, what we think it means.
3. Two or three recommendations. For each, say what we would change, what we
would leave alone, and what we would check in two weeks.
4. One paragraph titled "What this report cannot tell you".
Rules: do not call any asset "bad" or "failing". Do not claim an asset caused
conversions. Do not invent figures. If an asset has fewer than 3,000
impressions, say it is too early to judge. Use British English.
CAMPAIGN TOTALS (use these for headline figures):
[paste from the campaign view]
DATA:
[paste cleaned export here]A realistic first output would say that the 15-second video and Image 3 carried the group, that Headline 5 spent about $290 for a single conversion, and that Image 6 spent $61 without a conversion. Here is where you intervene. The AI will often write "Headline 5 is underperforming and should be removed". Rewrite that as "Headline 5 ranks lowest among the headlines and has produced one conversion on $290; we will test a more specific replacement alongside it for two weeks before removing it". The difference is not politeness. Removing an asset can change how Google combines the rest, and a client who hears "remove" may quote it back to you later when the month dips.
The second correction is arithmetic. The AI may add the asset costs to a "total spend of $1,268". The real campaign spend comes from the campaign view, and as Google notes, asset sums may not match it. Tell the model to quote only campaign-level totals that you supply, and paste them in a separate block labelled as shown in the prompt above.
Turning the summary into something a client acts on
A client report earns its keep when it ends with decisions the client can make, not a list of observations. For each recommendation, make the AI produce three fields: the change, the reason in one sentence tied to a figure, and the review date. A filled-in version for the gift shop might read: "Change: add two new headlines that name a price point, such as 'Hampers from $35'. Reason: 'Gifts for every budget' ranks lowest of the headlines and the best-ranked headline mentions delivery speed. Review: two weeks after the new headlines go live, comparing conversions per $100 spent across the headline group."
Notice what that wording avoids. It does not say the new headline will win, and it ties the review to the date of the change rather than to a calendar month. Google's optimisation tips for Performance Max allow one to two weeks, or at least one full conversion cycle, for a campaign to settle after a significant change, so a review at a fixed offset from the edit is fairer than a review on the first of the month. Our PPC audit template uses the same idea: dated checks, not vague impressions.
Four more client situations and how to word each one
A strong label on an asset that costs too much
A sports club's Performance Max campaign promotes junior memberships. One image is labelled Best and has the most impressions, but its cost per conversion is double the group average (illustrative). The client's instinct is to add more images like it. The honest explanation is that the label compares the image with its siblings; it is not a verdict on what each membership costs, and the image's own cost per conversion is one of the ratios Google calls directional. Ask the AI to rewrite the paragraph so it separates "most effective by Google's ranking" from "cost per conversion", and to propose a test: run one new image that borrows the subject but changes the offer text on the image.
A client who wants every Low asset removed
A coworking space owner sees four Low labels and wants them gone. Use the AI to draft a short explanation: Low is relative, so some asset must always rank lowest; replacing all of them often means the next four rank lowest in their place. Offer a rule instead: remove an asset only when it has had enough traffic to be ranked, has been Low across two reporting periods, and a replacement is ready. The AI is good at producing that kind of two-sentence, jargon-free justification when you give it the rule.
Learning labels that look like neglect
An event venue added six new videos on the 20th. On the 28th, the report shows five labelled Learning. A client reading "Learning" may assume nothing is happening. Draft a standing line the AI can reuse: "New assets show as Learning until they have gathered enough data to be ranked; we judge them after two full weeks." Put it in your report template so it appears every time there is a new asset, and the conversation stops being a surprise.
An Ad Strength rating the client wants pushed to Excellent
Google's Ad Strength help describes the Performance Max ratings by what the asset group contains. Poor means only the minimum assets, so the group serves to some inventory; Average means every asset type needed for all inventory; Excellent means optional assets too, such as sitelinks, a horizontal logo and a square image. Google says a higher Ad Strength helps maximise performance, but the rating describes the setup, not the results. A client who sees "Good" may ask for more assets purely to turn the bar green. Fill the gaps that are cheap (a missing logo format, sitelinks, a second video), since Google's asset-level metrics guidance suggests aiming for Excellent before replacing or refreshing assets, but write plainly that the score measures setup breadth and the report's performance figures are what you will judge by.
Checking the AI's draft before it goes out
Treat the AI as a fast first drafter and yourself as the editor. A short checklist catches the usual failures in this kind of summary.
- Every figure in the text appears in the data you pasted. Search for each number.
- No sentence says an asset "caused", "drove" or "won" conversions.
- No campaign total was computed by adding asset rows.
- Assets with little traffic are described as too early to judge, not as poor.
- Recommendations name a change, a reason and a review date.
- The "cannot tell you" paragraph survived and mentions the per-asset counting and, if used, the relative labels.
- Nothing in the text reveals customer details or another client's results.
If the model keeps adding figures you did not supply, move the instruction "use only figures in the data" to the end of the prompt as well as the start, and reduce the data to the columns that matter. Long exports with twenty columns invite invention.
A repeatable monthly rhythm
The report only becomes cheap when the steps are fixed. A workable cycle for an agency running several Performance Max clients is: export on the first working day, clean columns in a spreadsheet template (a saved view or script does this in seconds), paste the cleaned data and campaign totals into a saved prompt, edit the draft, and send with a three-line summary at the top. Expect roughly fifteen to twenty-five minutes per client once the template exists, against the better part of an hour for writing from scratch (illustrative; time your own first three).
If you already automate the rest of your monthly report, slot this section into that flow rather than building a separate one; the wider approach is in how marketing agencies automate client reporting with AI. Asset ratings on a badly structured campaign explain very little, so if the account has not been reviewed recently, do that first.
Where asset reports stop and other reports begin
Clients often ask about things the asset report cannot answer: which search terms triggered the ads, whether brand queries are inflating results, or whether a poor week came from a seasonal dip. Those need different views. Search term reporting and the controls in AI Max negative keywords cover the query side, and a note in your report that points the client to the right section keeps asset commentary from being stretched to cover questions it cannot answer.
For local service clients there is a further change in motion: Google has said that Local Services Ads are moving into Performance Max in phases, and that old reports do not carry over. If a client in that group asks why numbers look discontinuous, the Local Services Ads export checklist explains what to save before the switch.
Where the explanation can mislead without anyone meaning it to
Four traps recur. The first is false precision: three decimal places of cost per click on an asset whose ratio Google calls directional. Round it, or leave it out. The second is survivorship: reporting only the Best assets so the client feels good, while the Low ones, which cost money, vanish from the page. The third is attribution language. "This image drove 11 conversions" is not what the report says; it says the image was served in ads that recorded those conversions. The fourth is stale advice: the interface and labels change, and Google's help pages are revised. Reread the relevant help page when you next update your template, and note the date you checked it in the template's footer.
One practical safeguard is to keep a one-page "how to read this report" insert that stays the same every month. It lists the caveats in the table above in client language and takes the argument out of each individual report. Draft it once with AI, edit it with a colleague, and have the client see it in the first report so that later months can refer back to it.
A three-line cover note the client will read
Most clients read the first screen of a report and nothing else, so ask the AI for a cover note after the full section is edited, and feed it the edited text rather than the raw data. That way the note can only repeat things you have already checked. For the gift shop it might come back as: "Hampers brought in most of this month's conversions, led by the table photograph and the delivery headline. One headline cost $290 for a single conversion, so we are testing a replacement for two weeks rather than deleting it. Three newer assets are still too early to rate; we will report on them in December." Three sentences, three facts, one action, one date.
Read the note aloud once. If any phrase would make the client ask "what does that mean?", the AI has smuggled in jargon, and it is worth adding "no ad-platform terms the client has not used themselves" to the prompt. Keep the corrected version as your standing example in the prompt so the next draft starts nearer the target.
Done well, the AI does the tedious reshaping and you do the judgement. The client gets a short, honest explanation they can repeat to their own team, and you get a report you are comfortable defending in a review call.
Questions clients and account managers ask about asset reports
Can I tell a client that a Low asset is bad?
No. Google describes its performance labels as relative rankings against other assets of the same type, so some asset always ranks lowest. Something can be Low and still earn conversions. Say it ranked below its siblings, then look at conversions and spend before suggesting any change.
How long before I act on a rating?
Google gives no threshold for single assets, though its Performance Max tips allow one to two weeks, or a full conversion cycle, after a significant change. Set your own rule, such as real impressions plus one reporting cycle after any edit, write it into the report and tell the client in advance so a quiet week is not read as failure.
Should the AI see client conversion data?
Send only what the summary needs: asset text, labels, impressions, clicks, cost and conversions. Leave out customer names, order details and anything personal. Check the client contract and the AI tool's data settings first, and use a business-grade plan rather than a free personal account.
Why don't asset numbers add up to the campaign total?
Google's asset-level metrics are counted per asset served in an ad, so one impression that shows three assets registers three times. Summing assets overstates the campaign. Quote campaign and asset group totals as the headline and treat asset figures as directional.
Further reads
- AI Ad Copy Generator for PPC Agencies: Responsive Search Ads at Scale — Write the replacement headlines once an asset is retired.
- Best AI Tools for Advertising Agencies: Copy Variants, Mock-Ups, Reporting — Compare tools for mock-ups, copy and reporting in one place.
- Can AI Manage Your Google Ads Without an Agency? — See how far AI can run Google Ads before an agency is needed.
- Plumbing Google Ads: What to Bid On, What to Block and How AI Helps — A trade-specific view of what to bid on and block.
- AI Tools for Investment Managers: Client Reports and Fact Sheets — Another client-report workflow with strict wording rules.
- AI Media Planning for Small Agencies: Channel Mix and Budget Splits — A prompt for a first-draft media plan, a worked $30,000 split, and five checks that catch bad sums, false sources and channels that cannot spend.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Google Ads Help (About asset reporting in Performance Max; About asset-level metrics; About asset group reporting; About Ad Strength; Optimisation tips for Performance Max; About asset reporting for App campaigns; the Local Services Ads migration page); Google Ads API release notes (v23) and the AssetPerformanceLabel reference.