Search Console API and AI Analysis with Claude and ChatGPT

#01API options

From Google's Search Console API usage limits:

#02AI analysis

See Claude CSV analysis of GA4 and Ads data, Claude MCP connectors for marketers, ChatGPT data analysis of marketing exports and Analysing GA4 exports with Claude and ChatGPT.

Our view: AI-powered configuration inside Search Console (lesson 4) is for setting up views; external assistants are for interpreting exports.

#03Prompts for analysing exports

Always check AI output against the data. Assistants can misread columns or round numbers.

#04Troubleshooting

#05Using the Search Analytics API without code

Google's Search Analytics query reference documents every field, and you can try requests from Google's API documentation after signing in. A typical request body:

{
  "startDate": "2026-08-01",
  "endDate": "2026-10-31",
  "dimensions": ["page", "query"],
  "rowLimit": 25000,
  "dimensionFilterGroups": [{
    "filters": [{"dimension": "page", "operator": "includingRegex", "expression": "/blog/"}]
  }]
}

Per the reference, rowLimit accepts 1 to 25,000 (default 1,000), and startRow pages through further rows. Operators include contains, equals, notContains, notEquals, includingRegex and excludingRegex (RE2). Dates are in Pacific Time. Set dataState to all to include fresh, incomplete data, and type to discover, googleNews, news, image, video or web (the default). The API returns top rows, not a guaranteed complete set. Spreadsheet add-ons and Data Studio also call the API for you.

#06Reducing quota problems

• Request fewer dimensions at once; page and query together is the most expensive combination.
• Use shorter date ranges and loop over them.
• Cache results; don't re-request the same data daily.
• For big, ongoing needs, bulk export (lesson 18) avoids API quotas entirely.

#07A prompt library for Search Console exports

#08Verification rules

1. Ask the assistant to cite rows for every claim.
2. Spot-check three claims in Search Console.
3. Check arithmetic on any totals or percentages it reports.
4. Don't let it guess causes without evidence; treat suggested causes as hypotheses.
5. Keep the prompt and the export date so next month is comparable.

#09Data handling

Search Console exports contain URLs and queries. Queries can occasionally include personal information people typed into Google. Remove any such rows, follow your company's policy for AI tools, and prefer workspace or enterprise accounts with data controls when your company provides them.

#10Ledgerline walkthrough: a cannibalisation check

1. The SEO lead pulled page and query data for 3 months with the API Explorer, filtered to the blog.
2. They uploaded the CSV to ChatGPT and used the cannibalisation prompt.
3. ChatGPT listed queries where two blog posts competed. The lead verified three in Search Console with a query filter and the Pages tab.
4. Two posts covering "invoice approval workflow" were merged, with a redirect from the weaker one. The date was annotated.

#11Questions practitioners ask

Is the API data the same as the UI? It uses the same data, but returns top rows and has row limits, so totals can differ slightly.

Can I use the API for the generative AI report? Check the current API reference; don't assume a report in the UI has an API endpoint.

Should I use AI-powered configuration or an external assistant? Use AI-powered configuration to set views inside Search Console; use an external assistant to analyse exported data.

#12Paging through results

For large sites, loop with startRow, increasing it by your rowLimit until a response returns no rows. Keep dimensions minimal; each extra dimension multiplies rows.

#13Analysis questions by goal

Paste exported, non-sensitive aggregate data and ask specific questions:

• "Here are 12 months of clicks by section. Which sections changed most, and which months?"
• "Group these queries into themes and list the top themes by impressions."
• "Compare these two date ranges and list pages with the biggest click drops and their impression change."

Check every claim against the data. Assistants can miscount or invent patterns.

#14API errors and fixes

#15Where this fits in the monthly routine

Use the API when the UI limits get in the way: more rows, scheduled pulls into a sheet, or many properties. Keep the pull small and documented so someone else can run it.

#16Quick checks before you move on

• You can run a query with a regex filter and paging.
• Credentials are stored securely and not shared in sheets or chat.
• Every assistant claim about the data has been checked against the export.

#17Worked example: Ledgerline's AI analysis

Ledgerline is a fictional accounts-payable software company. Details and numbers are illustrative.

Ledgerline's SEO lead exported two months of page data and asked Claude for top decliners. Claude flagged the integrations section; checking Search Console confirmed it. One suggested cause was wrong, so the lead noted it and kept the prompt.

Walkthrough

1. Export the Performance report for two periods (Pages tab, with Compare).
2. Remove any columns or rows containing personal data.
3. Upload the CSV to Claude or ChatGPT.
4. Ask for the top 10 decliners with likely reasons, citing rows.
5. Check three of the claims in Search Console yourself.
6. Save the prompt for next month.

#18Try it yourself

Export one month vs the previous month by page, and ask an AI assistant to identify the top decliners and group them by section. Verify the result against Search Console.

#19Quick quiz

1. Does the Search Analytics API have quotas?

Answer: Yes, load quotas over 10 minutes and a day, plus per-second, per-minute and per-day request limits.

2. Does URL Inspection API share that quota?

Answer: No. It has its own per-site and per-project quota.

3. Should you trust AI summaries without checking?

Answer: No.

4. What should you remove before uploading exports?

Answer: Personal data.

#20What should you check before the next lesson?

• You know the API quotas.
• You tested one AI analysis prompt.
• You verified its findings.

Next: Search Console and AI visibility beyond Google. Or return to the Search Console for Marketers track.

#21Common questions

Do I need to code to use the API? Not necessarily; connectors and add-ons exist, but code gives more control.

Can Claude connect to Search Console directly? Via MCP connectors where available; see the Claude track.

Is it safe to upload exports? Follow your company's data policy and remove personal data.