Documentation
Reports AI is Adnine's flexible reporting workspace for Google Ads. Use it when a standard dashboard does not answer the exact question you have: describe the table, metric, date range, comparison, sorting, and output format you need, then review the generated result and download it as a CSV.
- Best for
- Custom Google Ads analysis and repeatable reporting
- Input
- Natural-language prompt
- Data source
- Google Ads metrics and dimensions available to your Adnine account
- Output
- Generated table with CSV download
- Team feature
- Saved and shared Prompt Library
#What Reports AI does
Reports AI creates custom reporting tables from plain-language instructions. Instead of choosing from a fixed report template, you tell Adnine what you want to analyze: the Google Ads entity level, the metrics, the dimensions, the time period, any calculations, and the way the final table should be sorted or formatted.
Use it for ad hoc questions, recurring weekly reports, client-ready exports, QA checks before account reviews, and analysis that needs calculated fields such as weighted averages, growth rates, period deltas, or custom column names.
Common jobs
Rank performance
List campaigns, ad groups, keywords, search terms, locations, or devices by ROAS, CPA, conversions, cost, CTR, or another metric.
Compare periods
Show one date range beside another and add absolute or percentage-change columns.
Calculate custom metrics
Request weighted quality scores, conversion rate by match type, spend per city relative to conversions, or CTR drops by device.
Prepare exports
Control row limits, sorting, column names, number formats, and CSV compatibility before downloading.
#How a good prompt is structured
The best prompts are specific about scope, metrics, dates, calculations, row count, and output format. If any of those details are missing, Reports AI may still return a useful table, but the result is easier to review when the prompt already defines the structure you expect.
Prompt building blocks
Use these parts together when you want a reliable, reusable report.
| Prompt part | What to specify | Example |
|---|---|---|
| Entity level | Choose the level of the table. | Campaign, ad group, keyword, search term, device, city, or account. |
| Date range | Name the reporting window and, when needed, the comparison window. | Last 30 days vs. the previous 30 days. |
| Metrics | List the Google Ads metrics and calculated values you want. | Cost, conversions, CPA, ROAS, CTR, impressions, weighted quality score. |
| Dimensions | Add the breakdowns that explain the metric. | Device type, campaign, ad group, keyword match type, location. |
| Calculations | Describe formulas, weighting, deltas, ratios, and aggregation rules. | Weight quality score by impressions before aggregating to ad group. |
| Output control | Set row count, sorting, column names, and CSV formatting. | Top 50 rows, sorted by CPA descending, with European decimal commas. |
Prompt patterns that work well
| Need | Prompt pattern |
|---|---|
| Rank top or bottom performers | Return the top 50 campaigns by ROAS over the past 30 days, sorted descending. |
| Find weak performers | Return the lowest 100 ad groups by conversion rate, excluding rows with fewer than 100 clicks. |
| Compare channels or devices | Show clicks, cost, conversions, and CTR by device type for the last 30 days compared with the previous 30 days. |
| Prepare an export | Use European comma notation, double-quote every cell, and name the columns Campaign, Cost, Conversions, CPA, and ROAS. |
| Create a repeatable summary | Generate a weekly campaign summary with spend, conversions, CPA, ROAS, largest changes, and the campaigns that need review. |
#Google Ads data access
Reports AI fetches Google Ads data for the accounts available in your Adnine workspace. The prompt can request standard Google Ads metrics and dimensions, then ask Adnine to aggregate, compare, or transform them into the table you need.
Data you can ask for
The original guide describes broad access to Google Ads metrics and dimensions. Use the exact fields your team uses in reporting so the generated table is easy to validate.
| Data type | Examples | Typical use |
|---|---|---|
| Performance metrics | Impressions, clicks, cost, conversions, CTR, CPA, ROAS. | Rank performance, find outliers, and compare efficiency. |
| Entity dimensions | Account, campaign, ad group, keyword, search term. | Choose the reporting level and identify where action may be needed. |
| Context dimensions | Device type, location, date, match type. | Break performance into the segments that explain movement. |
| Calculated fields | Weighted averages, absolute change, percentage change, ratios. | Turn raw metrics into decision-ready columns. |
#Period comparisons
Reports AI is useful when a single date range is not enough. Ask it to place two periods side by side, calculate absolute change, calculate percentage change, and sort by the column that matters most.
Comparison prompt options
| Comparison | Use when | Prompt detail to include |
|---|---|---|
| Last period vs. previous period | You want recent movement without seasonality adjustments. | Last 30 days, previous 30 days, absolute change, percentage change. |
| Week over week | You are preparing a weekly team or client update. | This week, previous week, sort by conversion change or CPA change. |
| Month over month | You need management reporting or pacing analysis. | Current month to date, previous month to date, include spend and conversion value. |
| Custom windows | You are checking before and after a campaign change. | Use the exact start and end dates for both periods and name the change you are evaluating. |
For example: Generate a table with ad group name, conversions last 60 days, conversions in the 60 days before that, percentage change, and absolute change. Sort by absolute change descending.
#Sorting and row controls
Add row and sorting instructions directly in the prompt. This keeps large accounts manageable and makes the output easier to review before export.
Useful table controls
| Control | What it changes | Example prompt phrase |
|---|---|---|
| Row limit | Controls how many rows Reports AI returns. | Return the top 20 campaigns or the lowest 100 ad groups. |
| Sort column | Determines which result appears first. | Sort by spend descending, CPA descending, or conversion rate ascending. |
| Minimum volume | Prevents low-data rows from dominating the report. | Only include rows with at least 100 clicks or at least 10 conversions. |
| Filters | Limits the report to the scope you care about. | Only include Search campaigns, brand campaigns, mobile traffic, or one country. |
| Column order | Makes the table match your reporting template. | Put campaign name first, then cost, conversions, CPA, ROAS, and change columns. |
#Prompt Library and team sharing
The Prompt Library stores prompts that your team expects to reuse. Save prompts for recurring weekly reporting, client review templates, QA checks, or analysis patterns that multiple teammates need to run consistently.
What to save
Recurring reports
Weekly campaign summaries, monthly client exports, pacing checks, and account-health snapshots.
Specialist analysis
Weighted quality score, search-term cleanup support, device performance, location efficiency, and match-type reviews.
Formatting templates
CSV-ready prompts with agreed column names, row limits, decimal format, and sorting.
Team standards
Shared prompts that make reports consistent across account managers, analysts, and client teams.
When a prompt becomes part of your team process, name it clearly and include enough context that another teammate can run it without rewriting the instructions. A good library entry explains the purpose, the intended scope, the reporting period, and any output formatting expectations.
#Scheduled reports
Scheduled reports run a Report AI prompt automatically on a recurring cadence, save each result to a history you can browse, and email the people who need it when the report is ready. Open the Scheduled Reports page under the AI menu to manage everything in two tabs: History for past runs and Schedules for the prompts you have set to run.
Each schedule targets one Google Ads account and runs one prompt. Write the prompt directly, or pick a saved prompt from your Prompt Library, so a report you already trust becomes a recurring deliverable without rebuilding it. A daily process checks every active schedule and runs the ones that are due; paused schedules are skipped, and Run now still works on demand.
Cadence options
Set how often a schedule runs when you create or edit it.
| Frequency | What it does | What you choose |
|---|---|---|
| Daily | Runs once every day. | Nothing else to configure. |
| Weekly | Runs on the specific weekdays you select. | One or more days from Monday to Sunday, such as Tuesday only, or Monday, Wednesday and Friday. |
| Monthly by day of month | Runs on a fixed calendar day each month. | A day number from 1 to 31; if a month is shorter, the run is clamped to the last day of that month. |
| Monthly by weekday rule | Runs on a relative weekday each month. | A position (first, second, third, fourth, or last) and a weekday, such as the first Monday or the last Friday of every month. |
- Open the Schedules tab Go to Scheduled Reports under the AI menu and switch to the Schedules tab, then choose New schedule.
- Set the prompt and account Write the report prompt or pick one from your Prompt Library, give the schedule an optional title, and choose the single Google Ads account the report runs against.
- Choose the cadence Pick Daily, Weekly with the weekdays you want, or Monthly by a day of the month or by a weekday rule such as the first Monday.
- Add notify recipients Enter the email addresses that should be notified when each run completes. Leave it empty if you only want the run saved to history.
- Activate and save Keep the schedule active so it runs automatically, or pause it to keep it without running. You can run any schedule immediately with Run now.
The History tab shows every past run as a tile with the prompt, the run status, the date and time, the account it ran against, and the number of rows returned. Open a tile to load that run's full result in the standard report table, where you can search, filter, sort, fit columns, expand the view, and download the result as a CSV, exactly like a report you generated by hand.
A schedule is the fastest way to make a proven report recurring. Save the prompt in your Prompt Library, point a schedule at it, and the same report arrives on cadence with its run history kept for you and your team.
#CSV output and downloads
Every Reports AI result is returned as a table and can be downloaded as a CSV file. Use the prompt to decide how the exported file should behave in Excel, Google Sheets, Looker Studio, or your client reporting workflow.
CSV controls
| Output choice | Why it matters | Prompt phrase |
|---|---|---|
| Number format | Keeps decimals and currency readable in your region. | Use comma as decimal separator or use dot decimals. |
| Quoting | Prevents commas, symbols, or text values from breaking CSV columns. | Enclose all values in double quotes. |
| Column names | Matches the naming used in a spreadsheet template or client report. | Rename columns to Campaign, Cost, Conversions, CPA, and ROAS. |
| Row count | Keeps exported files focused and reviewable. | Return the top 50 rows only. |
| Sort order | Makes the exported report open with the most important rows first. | Sort by CPA descending or by absolute conversion change descending. |
For example: Return the top 50 campaigns by CPA over the last 30 days. Use comma as decimal separator, and enclose all values in double quotes so they are CSV compatible.
#Review workflow
Review the generated table before using it in a client report or optimization decision. The table should match the requested account scope, date range, metrics, calculations, sorting, and row count.
- Write the first prompt Start with the table structure: entity level, metrics, date range, comparison period, row limit, sorting, and formatting.
- Inspect the generated table Check whether the rows, columns, date range, metric names, and calculated values match the request.
- Refine the prompt If the table is too broad, add filters. If it is hard to read, rename columns, change the row limit, or adjust the sort.
- Validate the interpretation Use Google Ads context and your business rules before turning the report into an optimization recommendation.
- Save useful prompts Add proven prompts to the Prompt Library and share them with teammates who need the same reporting pattern.
- Download the CSV Export only after the table has the right scope, row order, columns, and number format for the next workflow.
#Example prompts
Prompts you can adapt
These examples preserve the core use cases from the original Reports AI guide.
| Use case | Prompt |
|---|---|
| Weighted quality score by ad group | Generate a table with weighted quality scores per ad group. First get keywords, date, impressions, and historical quality score in the last 30 days. Then aggregate per ad group, weighting the quality score by impressions for each ad group. Return ad group name, weighted quality score, and impressions last 30 days. |
| Conversion change over time | Generate a table with ad group name, conversions last 60 days, conversions in the 60 days before that, percentage change, and absolute change. Sort the table by absolute change descending. |
| CSV format customization | Return the top 50 campaigns by CPA over the last 30 days. Use comma as decimal separator and enclose all values in double quotes so they are CSV compatible. |
| Device performance comparison | Show clicks, impressions, cost, conversions, CTR, and CPA by device type for the last 30 days compared with the previous 30 days. Include percentage change for clicks, conversions, and CPA. |
| Low performers with volume guard | Return the lowest 100 ad groups by conversion rate over the last 30 days. Only include ad groups with at least 100 clicks, and sort conversion rate ascending. |
#Best practices
- Be specific about the account scope, entity level, metrics, date range, and comparison period.
- Use row limits and minimum-volume filters so the report stays focused on meaningful data.
- Ask for both absolute and percentage change when comparing periods; each answers a different question.
- Sort by the metric that drives the decision, such as CPA, ROAS, cost, conversion change, or CTR drop.
- Define custom calculations clearly, especially when weighting by impressions, clicks, cost, or conversions.
- Name columns in the prompt when the CSV will be used in a spreadsheet template or client report.
- Save proven prompts in the Prompt Library and share them with the team for consistent reporting.
- Review the generated table before using it in recommendations, presentations, or optimization workflows.