Best platform to monitor competitor ads in AI assistants: what we use and why

A few months ago, clients started forwarding us screenshots. A sponsored result sitting above the answer in ChatGPT. An ad inside a Google AI Overview. The question was always the same: who is that, and are they showing up for our prompts?

Those screenshots are anecdotes. You cannot build a plan on one person's search from one afternoon. What you need is a repeatable capture across the prompts that matter to the business, with a record of who advertised, what they claimed and where they sent people.

Start with the prompts, not the tool

Before we pick software for this work, we write the prompt set. Ads in AI answers are triggered by the prompt, so tracking ads on prompts nobody cares about gives you tidy charts and no insight.

We split prompts into three groups:

• Category prompts ("best X for Y"), where buyers compare options
• Competitor comparison prompts, where a rival may bid on your brand's territory
• Problem prompts, where someone describes a need without naming a product

Most advertiser activity we have seen clusters on the first two. That is a pattern from our own work, not a universal rule, so check it against your category.

What the data should capture

For each ad, we want the title and description, the image or price shown, the advertiser's name and domain, the landing page URL and the prompt that triggered it. Then we want a time series so we can tell a brief test from a sustained push.

If a tool gives you a count of ads and nothing else, it will not change a decision. The landing page is the useful part. It shows whether a competitor is sending traffic to a comparison page, a pricing page or a product page, and that tells you what they think the buyer is deciding.

What we run client programs on

We use Promptwatch for this. Its Ads Radar captures sponsored ads inside ChatGPT Search, Google AI Overviews and Google AI Mode for the prompts we track. The record has the fields above, and the views suit agency reporting: ads per day, top advertisers, a By Ad table and a By Prompt table.

The By Prompt table is the one we open first. It lists each tracked prompt with an ad count, your average visibility on that prompt and the latest ad. When a prompt has several ads and your visibility is low, that is where a client's paid and organic answers are both losing ground.

Some practical facts we pass on to clients so nobody is surprised:

• Ads Radar does not cover Perplexity.
• It does not report on a client's own OpenAI campaigns. That data stays in OpenAI's ads manager.
• On brand plans it starts at Business ($579/mo). On the agency side it is included on every plan, from Kick-off ($199/mo).

Why we pair ads with organic data

Ads are a symptom. The cause is often that a competitor is more visible than your client on a commercial prompt, and someone decided to pay for even more of the space.

Looking at the ad alone, a client's first instinct is to bid. Looking at the ad next to their organic visibility, citation sources and crawler activity, the better answer is sometimes content. If the cited sources on a prompt are all comparison articles your client does not appear in, no bid fixes that.

This is why we prefer a platform where ad capture sits next to prompt tracking and citation analytics. Switching between a standalone ad tracker and a separate visibility tool means two exports and an argument about which numbers to trust. Promptwatch also has a Visibility Score, Share of Voice and Position in Answer per prompt, and we line those up with the By Prompt ad view.

A simple monthly routine

1. Refresh the prompt set with the client's sales team. Add the questions that came up on recent calls.
2. Open the By Prompt view and sort by ad count. Note the prompts with new advertisers.
3. For each new advertiser, record the landing page and the claim in the ad copy. Flag anything that contradicts the client's positioning.
4. Check the client's own visibility on those prompts. If it is weak, check citation analytics for the sources AI is pulling from.
5. Decide per prompt: bid, publish, fix a technical gap or leave it alone.
6. Send a short summary. One page, three findings, one recommendation.

What you should not promise

Ad monitoring shows a sample of what was served to a tracked prompt at a given time and place. It does not reveal spend, bids or targeting rules. It does not guarantee that the same ad appeared for a real customer. And it cannot promise that your client will be recommended, with or without paid activity.

We also tell clients that the ad formats in AI assistants are changing. Features that exist today may look different in six months, so we date every report and re-check what the platform covers before each quarterly review.

Who needs this and who does not

If a client sells in a category with active paid search competition, ad monitoring in AI answers is worth a monthly look. If the category is local services with little paid competition, spend the time elsewhere first: crawler access, citation sources and content gaps usually matter more.

For agencies, the plan structure helps. Ads Radar is on every agency plan, along with unlimited projects and prompts, 10 seats and the same crawler logs and visitor analytics, so one workspace covers the ad work and the organic work.

If you want to see what a competitor ad review would look like for your own prompts, ask us for a sample audit. We would rather show you real captures than describe them.