ChatGPT Campaign Analysis Workflow

Overview

ChatGPT Plus can dramatically accelerate Google Ads campaign analysis by ingesting exported reports and returning structured performance breakdowns, optimization recommendations, and actionable next steps — without requiring direct account access. Using the Projects feature, you can maintain persistent, client-specific context across multiple analysis sessions.

This workflow was developed and demonstrated during an internal review session on 2025-10-08, using ^[1] and ^[2] as live examples.

Core Workflow

1. Export Reports from Google Ads

For each client account, download the following reports (CSV or XLSX):

Campaigns report — impressions, clicks, CTR, CPC, conversions, cost
Ad Groups report — same metrics broken down by ad group
Keywords report — search term performance, match types
Negative Keywords report — existing exclusions
Search Terms report — actual queries triggering ads (found under Insights & Reports → Search Terms)

Naming convention: Rename files to include the client name and month before uploading (e.g., bluepoint-campaigns-oct-2025.xlsx). This prevents confusion when files accumulate across months.

2. Create a Client-Specific ChatGPT Project

In ChatGPT Plus, create a Project for each client's Google Ads account:

Click New Project in the left sidebar
Name it clearly: e.g., Bluepoint ATM Google Ads
Upload all exported reports via Add Files
Add any relevant context notes (e.g., "Page view goal removed from conversions on October 8, 2025 — historical data prior to this date is inflated")

Why Projects? Unlike a standard chat, a Project retains all uploaded files and prior conversation context. Each new chat within the project can reference previous reports, decisions, and notes. This makes month-over-month analysis coherent without re-explaining the account from scratch.

3. Prompt for Analysis

Start with a broad analysis prompt, then drill down:

Initial prompt example:

"Here are the campaign, ad group, and keyword reports for [Client] Google Ads for October 2025. Please analyze performance across all campaigns. Identify strengths, weaknesses, and your top optimization recommendations. Note: page views were removed as a conversion goal on October 8 — data before that date has inflated conversion counts."

ChatGPT will typically return:

Per-campaign performance breakdown (CPC, CTR, conversion rate, CPA)
Identification of high-spend / low-conversion campaigns
Bid strategy recommendations (e.g., switch to Maximize Conversions, adjust Target CPA)
Landing page and audience quality flags
Specific action items (pause underperforming ad groups, add negative keywords, etc.)

Follow-up prompts to consider:
- "What additional data would help you give better recommendations?"
- "Which campaigns should we pause or restructure first?"
- "What negative keywords would you suggest based on the search terms report?"

4. Act on Recommendations

Common outputs from this workflow and how to handle them:

Recommendation	Action
Conversion tracking includes non-lead actions (page views, scrolls)	Remove from Goals in Google Ads → Conversions → Summary; add as Observations only
Landing page points to homepage instead of dedicated page	Update destination URL in ad to a relevant, conversion-focused landing page
PMAX campaigns showing inflated or unclear conversions	Pause or restructure after 1 week of clean conversion data
High CTR but low conversion rate	Review landing page relevance, add trust signals, call extensions, lead forms
Very low CPA (suspiciously low)	Audit conversion actions — likely counting micro-conversions as goals
Negative keyword gaps	Download search terms report, identify irrelevant queries, add as negatives

5. Maintain the Project Over Time

Each month, upload new reports to the same project
Rename files by month so ChatGPT can distinguish time periods
Add notes about any account changes (bid strategy switches, new campaigns, tracking fixes) as plain text messages within the project
The project accumulates institutional knowledge about the account over time

Key Principles

ChatGPT does not need account access. Exported CSV/XLSX files contain all the data needed. This keeps client credentials secure and the workflow portable.

Speed, not magic. ChatGPT is performing the same analysis a skilled analyst would — it's just doing it faster. The value is in compressing hours of review into minutes, not in surfacing insights a human couldn't find.

Garbage in, garbage out. If conversion tracking is broken (e.g., counting page views as conversions), ChatGPT will flag the anomaly but cannot fix it. Always audit conversion goals before trusting performance data. See ^[3].

One project per client per channel. Don't mix Bluepoint Google Ads data with Bluepoint Amazon data in the same project. Keep contexts clean.

Conversion Tracking Note

A recurring issue surfaced during this workflow: page views, scroll events, and sessions being counted as conversion goals in Google Ads. This inflates conversion numbers and causes Google's bidding algorithms to optimize toward low-value actions.

Correct setup:
- Goals (counted as conversions): Form fills, phone calls, button clicks, email clicks, PDF downloads (if meaningful)
- Observations only (not goals): Page views, scroll depth, session starts

When fixing this mid-campaign, note the date of the change. Historical data will remain inflated; filter reports from the correction date forward for accurate analysis. See the ^[4] for the live example with Bluepoint ATM (correction date: October 8, 2025).

^[5]
^[6]
^[7]
^[8]
^[9]