Connect Yandex Metrica to ChatGPT

Windsor.ai gives ChatGPT access to all 238 Yandex Metrica fields: two tables of individual session and page-view rows that ChatGPT can count and aggregate when you ask, plus shared fields such as account, source and date parts:

• Sessions (136 fields): traffic source, UTM parameters, GCLID, ecommerce purchase and product data, device category, screen resolution, operating system, and offline call tracking (duration, missed status, tag).
• Page views (81 fields): URL, page title, referrer, HTTP error status, bounce flag, file downloads, and UTM parameters.

Yes. Windsor.ai streams last traffic source and UTM data alongside purchase revenue and product fields on the same session row, so ChatGPT can count how many sessions from each source ended in a completed purchase.

Yes. Every page view includes an HTTP error field and a download flag, so ChatGPT can surface URLs with elevated error rates or unusual download activity.

Yes. Windsor.ai sends offline call duration, missed-call status and call tag with the session’s traffic source and campaign data, so ChatGPT can link each call to the campaign that brought the visitor.

• Count sessions by operating system and device category for the last 30 days. Is mobile traffic converting at a different rate than desktop?

• Find sessions with a purchase where the last traffic source was organic search. What UTM campaigns, if any, were involved earlier in the session?

• List the pages with the highest bounce rate (using the notBounce flag) over the last 14 days, along with their referrer.

• Break down offline calls by talk duration and campaign for the last 30 days. Which campaigns generate the longest, most substantive calls?

Create it in Yandex OAuth while signed in to the Yandex account that owns the Metrica tag or has access to it:

1. Open oauth.yandex.com, create an app and choose For API access or debugging.
2. Under data access, add metrika:read. If you sign in with an organization account, also add passport:business.
3. Click Create app and copy the ClientID.
4. Open https://oauth.yandex.com/authorize?response_type=token&client_id=YOUR_CLIENT_ID with your ClientID, click Allow and copy the token shown.
5. Paste it into the Authentication Token field in Windsor.ai.

Windsor.ai only reads your Yandex Metrica data, so metrika:read is the only Metrica permission it needs. Yandex describes the same steps in its Metrica API authorization guide.

The Counter ID is your tag number. Yandex’s English interface calls counters “tags”. Open any report for the tag in Metrica: the number after id= in the address bar is the Counter ID. It is also in the tag code under Settings, on the Tag tab.

To report on several sites, enter each tag’s Counter ID with a token and click Add Account again.

The token has usually expired or lost access. Yandex issues tokens for a limited time (its documentation gives one year for Metrica access as an example). A token also stops working when the Yandex account’s password changes, or when its owner loses access to the tag. Create a new token with the same steps and add it in Windsor.ai with the same Counter ID.

Windsor.ai is free to start, no credit card required. Paid plans start at $19/month and scale by the number of sources and accounts. See the pricing page.

Yes. Windsor.ai supports 350+ data sources, so you can connect Yandex Metrica alongside Google Ads or Meta Ads to see how paid campaigns relate to on-site session behavior and purchases.

Yes. These pages cover the Yandex Metrica fields, the ChatGPT setup and web analytics questions to ask with AI:

• Yandex Metrica Field Reference: every session and page-view field ID with its type.
• ChatGPT Integration Documentation: installing the Windsor.ai plugin in ChatGPT, step by step.
• How to use Windsor.ai for web analytics with AI: example questions that combine analytics, Search Console and ad data.