How to Connect Google Search Console to Codex With Aident Loadout

To connect Google Search Console to Codex, use Aident Loadout to discover the current Search Console Actions, authorize your Google account through Vault, select the exact verified property, and begin with a small read-only query. You do not need to run a local MCP server, paste a client secret into a prompt, or give the agent permission to change a sitemap.

The useful outcome is not a large export. It is a repeatable brief that compares equal complete periods, preserves the query and page dimensions, separates brand demand from broader discovery, and turns each finding into one reviewed action.

What You Need Before Connecting

Prepare four things:

• a verified Search Console property;

• a Google account with permission to read that property;

• the public Aident CLI installed and authenticated; and

• a decision you want the data to support, such as which existing page to improve or which missing intent deserves a new canonical.

Give Codex the canonical setup instruction:

Follow https://aident.ai/SETUP.md

Then confirm the account and Vault state:

aident account auth status
aident vault vault --action status

Ask Codex to find the current Google Search Console integration and inspect the live input schema before execution. Do not save an internal Action identifier in a permanent prompt. Catalog versions and accepted inputs can change.

Step 1: Connect Through Vault

Ask Codex to start the Google Search Console connection through Aident Loadout. Complete the Google authorization flow in the browser, then return to the agent and confirm that the connection is ready.

Keep authentication separate from analysis. Do not paste OAuth tokens, cookies, client secrets, or exported credentials into the conversation or repository. Vault should hold the connection while Codex receives only the permission needed to execute the selected Action.

If the account can access several sites, list the available properties and choose the exact canonical property. A domain property and a URL-prefix property can expose different scopes. Record the selected property in the brief so another analyst can reproduce the query.

Step 2: Prove the Connection With a Small Read

Start with a bounded, read-only Search Analytics query. Request a short date range, a small row limit, and the query and page dimensions. This verifies the account, property, schema, and result shape before a larger export.

Expected result:

{
  "property": "sc-domain:example.com",
  "dateRange": {
    "start": "YYYY-MM-DD",
    "end": "YYYY-MM-DD"
  },
  "dimensions": ["query", "page"],
  "rows": [],
  "dataState": "final"
}

If the request fails, check the property identifier and account permission before changing the query. A successful Google login does not prove that the selected account can read every property.

Step 3: Compare Equal Complete Periods

Search Console data usually lags by two or three days. Choose the latest complete end date, then compare it with an immediately preceding period of the same length. Keep the property, search type, dimensions, filters, and data state identical.

For example, this article's research used two equal 28-day windows:

This is a first-hand workflow check for Aident's site, not a performance benchmark for other sites. It also does not prove that the additional search traffic created new users or revenue. This attribution workflow connects content visits to PostHog product events.

Search Analytics prioritizes top rows rather than guaranteeing a complete export. The API accepts a rowLimit of up to 25,000 and a startRow offset, while the documented export ceiling is 50,000 rows per day per search type. Paginate when the result reaches your row limit, but label the result as privacy-filtered and top-row data rather than a complete keyword ledger.

Step 4: Split Brand Demand From Broader Discovery

A total traffic increase can hide whether new people are finding the site. Apply one documented, case-insensitive brand rule to both periods. Keep the total and filtered views together so the filter remains auditable.

For the same two windows, a directional filter excluding explicit aident, aiden, and loadout query strings produced:

This is not a precise non-branded taxonomy. Misspellings, product names, mixed-intent queries, and anonymized queries can remain. Use the split as a directional diagnostic, preserve the regex in the brief, and review the highest-impact rows manually.

Step 5: Turn Rows Into Page Decisions

Group findings by page and assign each one to a small decision set:

Do not create a new article merely because a query has impressions. First search the repository for an existing owner. In this run, a query about Codex model capacity had 25 clicks and 786 impressions at an average position of 8.06, with a close variant adding 6 clicks and 252 impressions. An existing capacity guide already owns that intent, so the correct decision is to refresh that canonical rather than create a competing page.

For every recommendation, preserve the query, landing page, current and prior metrics, position, intended reader, canonical owner, and next review date. Rank the queue by expected decision value, not raw impressions alone. A page near the first page with clear product relevance may deserve attention before a larger informational query that cannot lead the reader to a useful next step. The brief should also state what evidence would reverse the decision.

This research guide explains how Aident combines Search Console, product evidence, and canonical ownership.

Step 6: Inspect Only the URLs That Need Diagnosis

Search Analytics shows demand and performance. URL Inspection answers a narrower question: how Google sees one indexed URL. Use it only for selected pages with an indexing or canonicalization concern.

The documented URL Inspection quota is 2,000 requests per day per site and 600 requests per minute. A request returns the status of the indexed version, not a live crawl. Build a short review list from the Search Analytics findings, inspect those URLs, and keep sitemap submission or other write operations outside this read-only workflow.

Common Failure Modes

The Agent Uses the Wrong Property

List the properties, record the exact selected identifier, and run a small query before the full comparison. Do not assume the first property covers the intended hostname and protocol.

The Period Includes Incomplete Data

End the comparison on the latest complete date and store the extraction time. Reusing “the last 28 days” on different weekdays can silently change the evidence window.

Totals Are Treated as a Complete Keyword Export

Search Console can omit anonymized queries and prioritizes top rows. Preserve totals and dimensioned rows separately, paginate bounded requests, and disclose the limitation.

Search Traffic Is Reported as Product Growth

Search clicks and impressions measure acquisition to the site. Use tagged links and product analytics to measure setup starts, integration connections, activations, or revenue separately.

A Read-Only Audit Turns Into a Write

Keep sitemap submission and deletion out of the prompt. If a later workflow needs a write, inspect that Action separately and require explicit approval.

Reusable Codex Prompt

Paste this after setup and replace the bracketed values:

Use Aident Loadout to confirm my Google Search Console connection and inspect the current read-only Search Analytics schema. List the properties available to my connected account and ask me to select the exact property. Run a small proof query before the full export. Then compare the latest complete [period length] with the immediately preceding equal period using the same property, search type, query and page dimensions, filters, and final data state. Paginate within the documented limits. Preserve totals separately from privacy-filtered rows. Apply and disclose this brand filter: [brand regex]. Group findings into protect, improve, expand, or enter, and check the repository for an existing canonical before recommending new content. Do not submit or delete a sitemap, perform another write, or expose credentials.

Set up Aident Loadout and connect Google Search Console to Codex.

Sources

• Google Search Analytics query method

• Google Search Analytics overview

• Google guide to getting all available Search Analytics data

• Google Search Console API usage limits

• Google explanation of Search Console data freshness

• Google explanation of anonymized queries

• Aident Loadout setup

Refresh this guide when the Search Analytics or URL Inspection schema, Google quotas, Aident Loadout connection flow, or linked measurement canonicals change.