# AI analyst for advertising data: what to check first

> An AI analyst for advertising data answers plain-language questions about your ad accounts. What it does well, and seven checks before trusting it.

AI in operations · The Intellsys team · 18 November 2026 · 6 min read

An AI analyst for advertising data is software you question in plain language about your connected ad accounts. It queries the data, shows how it reached the answer, and returns a figure, a chart or a file. It saves the hours spent exporting and joining reports. It does not remove the need to check: a fluent answer can rest on the wrong dates, the wrong accounts, or a metric each platform defines its own way.

## The short answer

- An AI analyst turns a plain-language question into queries across your ad accounts and returns the answer.
- Its value is time: no exporting, joining and rebuilding the same report every Monday.
- Check the dates, the accounts and the metric definitions before you repeat any number.
- Platforms count results by their own rules, so their figures do not simply add up.
- An analyst reads and explains. A person decides what to change in a campaign.

## What an AI analyst for advertising data does

A marketing head in a dealer-led business might run search, social and marketplace advertising across forty cities. The Monday question is simple: what worked last week? The answer usually takes someone half a day of exports. An analyst follows the same path, faster, and shows each step.

- **The question** "Which cities gave us the cheapest leads last week, on search and on social?"
- **The accounts** It works out which connected advertising accounts and which dates the question covers.
- **The queries** It queries each platform's data separately.
- **The join** It puts the results side by side, by city and by week.
- **The answer** It returns a table or chart, with the steps it took shown.
- **The check** A person reads the steps, checks the figures and decides what to do.

*How an AI analyst answers one question about advertising data, from the question to the person who checks it.*

The last step is not optional. The analyst is quick; whether its answer is right is still a person's job.

## Questions worth asking an advertising analyst

The most useful questions are the ones someone already answers by hand each week. A few examples from a dealer network, and the catch in each:

| The question | Why it helps | What to watch |
| --- | --- | --- |
| Which cities had the lowest cost per lead last month? | Moves budget towards cities that produce enquiries | Cost per lead says nothing about which leads became sales |
| Which campaigns spent most and produced fewest leads? | Finds waste quickly | A brand campaign may not be meant to produce leads |
| How did search compare with social for the new product launch? | One view across platforms | Each platform counts a lead by its own rules |
| What changed week on week, and why? | Saves rebuilding the same report | "Why" is a reading of the data, not a fact |
| Which marketplace listings drew traffic but few orders? | Points at listings to fix | Traffic and orders may be measured over different dates |

## What to check before trusting an answer

An answer in a confident sentence is easy to repeat in a review. Before you do, tick through these. Most take a minute, because a good analyst shows its working.

- [ ] **The dates** Confirm the period, and whether "last month" meant calendar month or the last 30 days.
- [ ] **The accounts** Check that every relevant ad account was included, and no test account.
- [ ] **The metric** Ask how "lead" or "conversion" was defined on each platform.
- [ ] **The money** Check the currency, and whether spend includes tax.
- [ ] **The join** See how cities or campaigns were matched across platforms with different names.
- [ ] **The steps** Read the reasoning the analyst shows, not only its conclusion.
- [ ] **One spot check** Compare one figure against the platform's own screen.

*Seven checks to make before repeating an AI analyst's answer in a review. A good analyst shows enough working to make each one quick.*

If the analyst cannot show its working, it is a guess with good grammar.

## Why platform figures do not simply add up

An error in a cross-platform answer is often not the AI's. It is in the data. Each advertising platform reports results by its own rules, and a customer who saw an advertisement on two platforms may be counted by both.

| What each platform reports | What the business needs to know |
| --- | --- |
| Results the platform credits to its own advertising | Leads the business actually received |
| Its own definition of a conversion | One definition, the same for every channel |
| Results counted over its own window of days | Results over the period the review covers |
| Clicks and form fills | Which of those became a visit, a quote or a sale at the dealer |

*Why adding up each platform's own figures overstates results. The right-hand column needs the business's own records beside the ad data.*

An analyst can show each platform's figures side by side. It should not quietly add them up as if they measured the same thing. What belongs on a view that joins these channels is in [the cross-channel marketing dashboard](https://intellsys.cloud/blog/cross-channel-marketing-dashboard).

## What an advertising analyst cannot tell you

Advertising data stops at the form. It knows that someone filled one in. It does not know whether the dealer called, whether the customer came in, or whether they bought. Unless the analyst can also see the business's own lead records, "which campaign sold the most" is out of its reach.

That join, from an advertisement to a sale at a dealer's counter, is its own problem, set out in [offline sales attribution](https://intellsys.cloud/blog/offline-sales-attribution). Until it is solved, treat cost per lead as a measure of enquiries, not of sales.

Nor should an analyst change a campaign. Whether software should ever act on what it finds, and at what level, is covered in [human in command AI](https://intellsys.cloud/blog/human-in-command-ai).

> **How Intellsys OS handles it.** AdGPT is the advertising analyst inside Intellsys OS. You question it in plain language about connected advertising data; it shows its reasoning step by step and returns charts and downloadable files.

## How Intellsys AI works as an analyst for advertising data

AdGPT is part of Intellsys AI. A person questions it in plain language about advertising data connected to Intellsys OS, which can include Google Ads, Analytics, Search Console, Business Profile and YouTube, Meta, LinkedIn, Amazon Ads, Seller and Vendor, and Flipkart. It shows its reasoning step by step, uses a sub-agent for each advertising platform, and returns charts and downloadable files. Work can be kept in projects with shared instructions, and it keeps a memory a person can read and edit.

Briefings arrive on a schedule, by email or WhatsApp. In Hybrid, AdGPT explains what moved and drafts the next plan, and marketing approves it. In AI Ops, agents shift budget and pause waste inside limits the business sets. More is on [the Intellsys AI page](https://intellsys.cloud/intellsys-ai), and the wider system in [what a business operating system is](https://intellsys.cloud/blog/what-is-a-business-operating-system).

## Questions people ask

### What is an AI analyst for advertising data?

It is software you ask questions in plain language about your connected ad accounts. It works out which accounts and dates the question covers, queries the data, puts the results together and returns a figure, chart or file, ideally showing each step it took.

### Can you trust an AI analyst's answer about ad performance?

Trust it after checking. Confirm the dates, the accounts included, how each platform defined a lead or conversion, and the currency. Read the steps it shows, and compare one figure with the platform's own screen before repeating the answer in a review.

### Why do figures from different ad platforms not add up to our leads?

Each platform credits results to its own advertising, by its own rules and over its own window of days. A customer who saw both may be counted twice. Compare each platform's figures with the leads your business actually received.

### Can an AI analyst tell which advertisement led to a sale?

Only if it can see your own lead and sales records as well as the ad data. Advertising data stops at the click or form. Joining it to what happened at the dealer needs the business's records, matched lead by lead.

---

[Request a demo](https://intellsys.cloud/demo) · [Blog](https://intellsys.cloud/blog)

This is the Markdown version of https://intellsys.cloud/blog/ai-analyst-for-advertising-data. Intellsys is built by GrowthJockey, a full stack venture architect.
