An agency portfolio review starts with one question: which client accounts need attention first, and why? Opening every account and running a full audit answers that question eventually. A shared comparison followed by a focused portfolio check gives you a practical review queue.
This guide covers the multi-client workflow in ChatGPT or Claude: account scope, comparable reporting, missing data, and prioritization. For team responsibilities and client reporting, use the AI PPC management guide for agencies. Once you choose an account to investigate, use the single-account Google Ads audit checklist or the weekly Google Ads review for its recurring checks.
Can ChatGPT or Claude review several Google Ads accounts together?
Yes, when you provide data for each account through a connection or exports. The assistant can organize a cross-account comparison and recommend where to investigate. A useful review preserves each client's identity, reporting dates, currency, conversion definition, and target; a combined spend total alone does not tell you which client needs help.
| Approach | What you provide | What to verify |
|---|---|---|
| Connected accounts through LoomaScale | Google Ads access and an explicit list of client customer IDs | Which accounts were actually read, reporting limits, and the dates returned |
| Separate account exports | Matching reports for each client, labeled with customer ID, dates, currency, and conversion definitions | Missing files, different filters, and whether recent data has changed since export |
The prompts below work as instructions in either ChatGPT or Claude. The connection and tools available in your assistant determine what it can retrieve. An export-based review can answer only what the supplied reports contain; it cannot inspect an account's current settings from a summary spreadsheet.
1. Confirm MCC access and the exact client accounts
A Google Ads manager account, formerly called My Client Center or MCC, lets you access multiple Google Ads accounts from one place. Google's manager account documentation describes cross-account reporting and account linking. The manager is the access route; the client customer accounts hold the campaigns you want to review.
For a connected workflow, follow the ChatGPT connection guide or the Claude connection guide. Connect the Google identity with access to the intended clients, then check which client accounts are enabled in LoomaScale. Selecting a default account for single-account work does not define the scope of a portfolio request: name the client IDs explicitly.
Start by reconciling the assistant's account list with your own client roster. Avoid relying on names alone when two accounts have similar names.
List the Google Ads client accounts available through this connection, with customer ID, account name, currency and time zone where available. My intended client roster is: [client name + customer ID for each client]. Match the available accounts to this list. Flag missing accounts, ambiguous names, manager accounts and unavailable metadata. Do not run the review on accounts outside this roster. Do not change anything.
A missing client needs an access check, not an inferred performance result. Check the Google login's access, manager-to-client link, account availability, and whether the client is switched on in LoomaScale. The pricing page explains the account allowance; a tool's batch size is separate from your plan's allowance.
2. Define a fair comparison before requesting rankings
Use explicit start and end dates for every account. Choose complete calendar days and a previous period of the same length, then note holidays, promotions, or other reasons the periods might differ. For accounts in different time zones, the same calendar dates still represent different UTC windows; avoid calling the data perfectly synchronized.
Give the assistant a small client brief alongside the account roster:
| Field | Why the review needs it |
|---|---|
| Customer ID and client name | Keeps findings attached to the right account |
| Currency and time zone | Prevents invalid money totals and explains reporting boundaries |
| Business objective and conversion definition | Separates purchases, qualified leads, calls, and other outcomes |
| Agreed CPA or ROAS target, if applicable | Compares each client with its own goal |
| Agreed budget for the period | Provides a basis for a pacing check; this is not implied by past spend |
| Tracking changes and conversion delay | Explains when reported results are incomplete or no longer comparable |
Do not rank an ecommerce account against a lead-generation account by raw conversions. Even two lead accounts may count different actions. Keep clients in meaningful groups and compare each with its own target and previous results.
For calculated metrics, use CPA = spend / conversions and ROAS = conversion value / spend, with value and spend in the same currency. If conversions are zero, CPA is undefined; if value tracking is absent, ROAS is unavailable. Ask for those labels rather than a misleading zero. For compatible combined results, calculate ratios from summed numerators and denominators instead of averaging account-level ratios.
Google's conversion lag guidance explains why recent CPA can look higher and ROAS lower before conversions arrive. Zero reported conversions is a reason to investigate tracking and traffic, not proof that the tracking is broken.
3. Compare performance across the client roster
LoomaScale's account comparison retrieves spend, impressions, clicks, conversions, and conversion value across accounts. It retains native currency and time-zone information and can add USD spend equivalents. Use those equivalents as approximate comparison figures, not historical exchange-rate accounting. If a conversion is unavailable, keep that account out of combined monetary totals and explain why.
The current comparison limit is up to 25 accounts per call. Requests can also return partially if an account is inaccessible, an allowance is reached, or the time budget is exhausted. Split larger rosters into explicit batches and reconcile the results against the original list.
Compare only these Google Ads client customer IDs: [IDs]. Current period: [YYYY-MM-DD] through [YYYY-MM-DD]. Previous period: [YYYY-MM-DD] through [YYYY-MM-DD]. Client goals, conversion definitions and agreed budgets: [briefs]. Retrieve each period separately using the account comparison tool. Batch at most 25 accounts per call. Show one row per client with customer ID, currency, time zone, spend, conversions, conversion value, CPA and ROAS where meaningful, and period-over-period changes. Keep native currencies visible. Label any USD equivalents as approximate. Do not sum mixed currencies or average account-level CPA or ROAS. Mark undefined metrics, missing goals and incomplete data explicitly. Reconcile the requested IDs with returned accounts. List skipped or unprocessed accounts and the reason; never treat them as zero-spend accounts. Do not change anything.
This produces the comparison layer. The assistant calculates ratios and changes from the returned data; the tool does not know the client's agreed targets unless you supply them. Check at least the flagged accounts against Google Ads using the same dates and conversion columns before making decisions.
4. Run a portfolio check and prioritize investigations
A portfolio audit adds a first pass over campaigns spending without conversions, expensive non-converting search terms, conversion-tracking signals, budget constraints, and disapproved ads. It helps select the next account to open.
LoomaScale's portfolio audit currently handles up to 15 accounts per call and returns 10 findings by default, with a maximum of 50. It uses bounded reads, including the top 25 search terms by spend per account, so it is not an exhaustive audit of every query or setting. A clean first pass does not establish that an account has no problems.
Set the non-converting-spend threshold deliberately. The default is your saved threshold, or 20 in each account's native currency. That is not the same economic threshold in USD, EUR, and UAH. For different currencies or client economics, use separate batches with suitable thresholds.
Run a report-only portfolio audit for these Google Ads client IDs: [IDs]. Use [start date] through [end date], with a non-converting-spend threshold of [amount] in each account's native currency. This batch shares that threshold; keep other currencies or thresholds in separate batches. Use batches of at most 15 accounts and request up to 50 findings per batch. State accounts audited, skipped accounts, findings omitted by limits, and any partial or truncated results. Combine the findings into a queue of accounts to investigate. For each: customer ID, issue, supporting evidence, affected spend in its currency, uncertainty, recommended next check, and owner. Use the client briefs to explain priority. Do not add overlapping campaign and search-term spend into a claimed savings total. Do not infer broken tracking from zero conversions alone. Do not change anything.
Prioritize an access or measurement gap when it prevents a reliable decision. Then consider delivery problems, budget constraints, and performance against the client's own goals. High spend can make an investigation urgent, but it does not make every flagged amount recoverable waste.
What should a multi-client review output look like?
Use an account-level queue with evidence and the next check. The following figures are illustrative, not client results or a LoomaScale benchmark. They show why the lowest CPA is not automatically the best account.
| Client | Observed result | Client context | Next check |
|---|---|---|---|
| A: lead generation | USD 1,200 spend; 12 qualified leads; CPA USD 100 | Target CPA USD 80; sufficient conversion time has elapsed | Investigate the 25% gap to its own target |
| B: ecommerce | EUR 2,000 spend; EUR 8,000 conversion value; ROAS 4.0 | Target ROAS 3.0; value tracking is verified | Check delivery constraints before proposing more budget |
| C: local service | USD 600 spend; zero recorded conversions | Conversion setup changed during the period | Verify measurement before judging campaign efficiency |
Client C needs a measurement check. Client A needs an efficiency investigation. Client B may warrant a budget discussion, subject to its business constraints. These are different decisions; a single ranking by spend or conversion count would hide them.
The completion note should say how many accounts were requested, compared in both periods, audited, and skipped. Keep an account with missing data in the queue until someone resolves it. A successful response covering only part of the roster is still an incomplete portfolio review.
5. Hand off each finding to the right client review
After the portfolio pass, investigate one client at a time. Carry forward the customer ID, dates, conversion definition, and evidence. The single-account audit checklist covers deeper diagnosis; the weekly review workflow covers recurring checks after the baseline is understood.
Keep any client-facing update limited to that client's account. Separate observations, hypotheses, proposed changes, and completed work. Assign an owner and a follow-up date to each open item. A report-only prompt defines the requested task; account permissions and integration controls determine what actions are technically possible.
Start with a small group of connected clients, verify the account coverage and calculations, then expand the same workflow in batches. Connect ChatGPT or Claude to Google Ads to build your first portfolio review with LoomaScale.
