Competitive research does not need a week. It needs one good question and a timer.

Most of us picture competitor analysis as a big project. A spreadsheet with forty columns. A slide deck nobody opens after the meeting. So we put it off, and we guess instead.

Here is what I keep noticing while I learn AI tools in public: the slow part of competitive research was never the thinking. It was the reading. Opening tabs, skimming homepages, scrolling through reviews. Claude can now do that part in minutes.

The thinking is still yours.

Below is the 15-minute workflow I put together, with the exact prompts, the free tools that fill in what Claude cannot see, and the one habit that keeps the results honest.

Study the site before you draw

Before marketing, I studied architecture. One of the first lessons is that nobody designs a building on a blank page. You start with site analysis, the research you do before design begins. Part of it is the neighborhood: identifying how the existing buildings around you are used. The neighbors shape what you can build.

Competitive research is site analysis for your marketing.

And just like on a real site, you do not need to measure everything. You only need to know what will change your design.

In my post on quantitative and qualitative data in market research, I wrote that you cannot market effectively to a blank wall. Your competitors are the walls next to yours.

The pressure is real, too. In Crayon's 2025 State of Competitive Intelligence report, a survey of more than 400 competitive intelligence professionals at B2B software companies, 55% said they had more competitive deals than a year earlier. The same report found that the number of CI professionals using AI grew 76% in a year. Dedicated CI teams are already doing this with AI. A solo marketer can do a small version of it before lunch.

What you need

All of these are free to start, as of October 2026.

  • Claude, with web search turned on. Web search is included on the Free plan. Anthropic's help center says you switch it on from the "+" button in the chat box (in the newest version of Claude there is no toggle, and Claude searches when it helps). It also says "every response includes citations, so you can easily verify sources yourself."
  • The Meta Ad Library. It lets you search "all the ads currently running across Meta technologies."
  • The Google Ads Transparency Center. Google describes it as a searchable repository of advertisers and the ads they have served on Google platforms, including Search, Display, Gmail and YouTube.
  • The LinkedIn Ad Library, if you work in B2B.
  • A timer. This matters more than it sounds.

If you pay for Claude Pro ($20 a month, or $17 a month billed annually), you can also use Research, which runs several searches that build on each other. It goes deeper, but each answer takes a few minutes and uses up your limits faster. For a 15-minute session, regular web search is enough.

The 15-minute workflow

Minutes 0–2: Write one question

Not "analyze my competitors." That is a topic, not a question. Write down one decision you are trying to make:

  • Why would someone choose them over me?
  • What are they promising that I am not?
  • What do their customers complain about that I could fix?

Then pick three competitors: one direct competitor, one bigger brand, and one that solves the same problem in a different way.

Minutes 2–7: Take the snapshot

Open a new chat with web search on and paste this. Starting with "Use web search" is the help center's own tip for making sure Claude actually searches.

Use web search. I am a [what you do] for [who you serve]. My competitors are [A, B and C] ([their websites]). Build a table with one row per competitor and these columns: who they target, their main promise in their own words, their offer and price, the proof they show (reviews, numbers, client logos) and their main call to action. Link a source for every cell. If you cannot find something, write "not found." Do not guess.

The last two sentences do the most work. When you give an AI tool permission to say "not found," you make it easier for it to admit a gap instead of filling it with something that sounds right.

Minutes 7–10: Look at their ads

Ad libraries are built for people to search and scroll, so this step is quickest by hand. Search each competitor in the Meta Ad Library, and in Google or LinkedIn if they advertise there. Screenshot two or three ads from each.

Drop the screenshots into the same chat. Claude reads JPEG, PNG, GIF and WebP images, up to 20 files per chat, and you can paste a screenshot straight from your clipboard. Then ask:

These are current ads from my three competitors. Which hooks, offers and calls to action repeat? What does every one of them say? What does none of them say?

That last question is where the useful answers live. What everyone says is the noise. What nobody says is your opening.

Minutes 10–13: Hear their customers

A competitor's website tells you what they want to be known for. Their reviews tell you what they are actually known for.

This is the Voice of Customer from my market research post: the exact words people use to describe their frustrations. Save those words. They are your next headline.

Search for public customer reviews and forum discussions about [A, B and C]. For each one, give me the three things customers praise most and the three things they complain about most, with a short quote and a link for every point. Only use pages you can open.

Minutes 13–15: Find the "so what"

Based only on what we found in this chat, where is the gap between what customers want and what these competitors offer? Give me one move I could make this week, and the evidence for it.

Then do the step most people skip. Click three of the links Claude gave you and check that each page really says what Claude says it does.

What this could look like

Here is a hypothetical, so you can see the pieces working together.

A freelance social media manager works with local restaurants. Her question: "Why would a restaurant owner hire one of them instead of me?"

The snapshot shows all three competitors promising more followers. Their ads repeat the same hook: a before-and-after screenshot of follower counts. Nobody mentions reservations, foot traffic or the owner's own voice. The reviews tell the rest of the story. Owners praise the pretty photos and complain that the captions sound like they could be about any restaurant.

The gap: everyone sells followers, and customers want posts that sound like the place.

The one move for this week: rewrite her homepage headline around what the owner actually wants, and post one example of a caption written in a client's voice.

Fifteen minutes. One question, one table, one move.

"Can I trust what an AI tool finds?"

Not blindly.

In a study published in March 2025, the Tow Center for Digital Journalism tested eight AI search tools on finding and citing news articles. Collectively, they gave incorrect answers to more than 60% of the queries. Claude was not one of the tools tested, but the lesson applies to every AI tool: the link is the receipt.

That is why every prompt above asks for a link and gives Claude permission to say "not found." And that is why you click before you act.

One more limit to keep in mind: Claude can only work with what is public or what you give it. It does not know your competitor's sales numbers, their email list or what their customers say on a private call. Treat the result as a first sketch of the site, not the final survey.

The honest part

Fifteen minutes gives you a snapshot, not a strategy.

Your competitor can launch a new offer tomorrow. One table will not tell you everything about a market. What it will do is replace guessing with a few facts you can check, and that is already more than most of us do before we write a campaign.

The real advantage comes from doing it again next month.

Make it a monthly habit

Save the workflow in a Claude Project. Projects are workspaces with their own chat history, files and instructions, and the Free plan allows up to five of them.

Put your four prompts and your list of competitors in the project. Next month, open it, paste the same question and compare. The changes between two snapshots often tell you more than either snapshot alone.

Try this today

Set a timer for 15 minutes. Pick three competitors. Write one question. Run the four prompts. Then choose one thing to do this week.

An architect does not study the building next door to copy it. They study it so their own building fits the site and still stands out.

What is the one question you would ask about your competitors?

Sources

If this was useful, the next piece will look at what to do with the gap once you find it: turning one competitive insight into a week of content. Subscribe so you do not miss it.