Connect Meta and Facebook Ads to ChatGPT

Choose the Meta ad account behind the results

Facebook and Instagram advertising managed through Meta belong to the same advertising integration. Connecting a personal Instagram profile or a Facebook Page alone does not establish access to the ad account that pays for the campaigns.

In the Adport project for this client, connect Meta, save the intended ad accounts, and enable agent access. Match an account ID to Ads Manager before starting a review. If several businesses use similar campaign names, ask ChatGPT to include the account ID in every table instead of relying on the display name.

The Meta integration guide describes provider prerequisites and supported capabilities. Access to a business's assets and permission to operate them remain separate from installing an AI plugin.

Install Adport and establish the conversation scope

Follow the ChatGPT MCP installation guide, select OAuth, and authorize the Adport project for this advertiser. Install the plugin and select Adport in the chat where you will review the campaigns. OpenAI explains how installed custom MCP plugins are selected in a conversation.

If the response lists another provider, restate Meta and the account ID. Treat the inventory as a scope check, not a recommendation to enable every discovered asset.

Compare campaigns with compatible outcomes

Choose one objective for the first comparison. A video-view campaign and a purchase campaign answer different business questions, even if one has a cheaper cost per result. Ask ChatGPT to show the definition behind the result column and identify metrics that the report does not supply.

Ask for the account's reporting basis and attribution settings when they are available. A period comparison can be useful without proving what caused the movement. Meta documents the Marketing API used for campaign management and reporting.

Ask a separate question about placements

A Meta account total is not automatically a Facebook-only result. Before asking which platform performed better, check whether the connected report supports the necessary placement breakdown. If it does not, request the available account view and identify the missing evidence.

Do not let ChatGPT invent a Facebook-versus-Instagram split from campaign names. Names describe an operator's intent; the requested report determines what delivery data was actually returned. A useful answer distinguishes placement, creative, audience, and objective instead of presenting one speculative explanation as a measured finding.

Check the budget level before previewing a change

When a review suggests a budget adjustment, establish whether the budget belongs to the campaign or an ad set. The same number can have a different effect at each level. Specify the exact entity ID, budget type, currency, and intended amount before requesting a preview.

Review the preview's old and new values and any policy adjustments. Approval to investigate Meta performance is not approval to change every campaign in the project. After an approved operation, request a fresh read of that entity and confirm the stored budget or status. Keep the evidence and decision together for the next review.

Resolve missing accounts and reporting mismatches

If the advertiser is absent, check business asset assignment, provider authorization, the saved Adport selection, and project scope. If account reads succeed but a particular report fails, preserve the error and identify the unavailable metric or breakdown rather than filling the gap with a guessed value.

For a dashboard mismatch, compare the exact account, dates, time zone, filters, objective, and conversion definition. Recent results can also change as reporting completes. Record which settings you reconciled before recommending a creative or budget action. Use the ChatGPT connection hub when the same review includes other advertising platforms.