Strategy & Planning

How to run competitive research with AI

Strategy & Planning4 AI tools7 steps6 friction points

Want an agent to handle this on your live data? See the Starch version →

Competitive research means knowing what your rivals are doing — their positioning, pricing, feature set, messaging shifts, hiring signals, and customer sentiment — before those changes affect your own pipeline. For most operators, it's a recurring obligation that never quite gets done properly. You mean to check competitor pages monthly, monitor their job boards, track their G2 reviews, and read whatever analysts write. The reality is a scattered folder of bookmarks and half-remembered conversations.

AI feels like the right tool here because competitive research is fundamentally a reading and synthesis problem. There's a lot of raw material — landing pages, LinkedIn posts, press releases, review sites, pricing pages — and very little of it requires human judgment to collect. The bottleneck is aggregation and pattern recognition across many sources, which is exactly what large language models are good at. If you can get the text in front of the model, the model can summarize, compare, and highlight what changed.

ChatGPT, Claude, and Gemini can all contribute meaningfully to competitive research today. Claude handles long documents well, making it useful for pasting in multiple competitor pages at once. Perplexity can run live web searches and cite sources, which helps when you want a current snapshot of a competitor's positioning. ChatGPT with browsing turned on can pull recent news and job postings. None of them connect to your business data — but for research that lives mostly on public web pages, they're genuinely useful starting points.

Skip the manual work → See the Starch version

Strategy & Planning4 AI tools7 steps6 friction points

AI walkthrough

How to do it with AI today

A practical walkthrough using ChatGPT, Claude, and other off-the-shelf LLMs — what they're good at, what you'll have to do by hand.

Tools that work for this

ClaudePerplexityChatGPTGemini

Step-by-step

1 Start with Perplexity: run a search for each competitor by name, asking for their current positioning, recent product announcements, and notable press coverage from the past 90 days. Save the cited sources.

2 Pull the full text of each competitor's homepage, pricing page, and 'About' page. Paste all of them into a single Claude session and ask it to extract their core value proposition, target customer, and any pricing structure it can infer.

3 Go to G2, Capterra, or Trustpilot and copy 15-20 recent reviews for each competitor. Paste into Claude and prompt it to identify the top three customer complaints and the top three praised capabilities for each product.

4 Open each competitor's LinkedIn company page and their job board. Manually copy current open roles into a document. Ask ChatGPT or Claude to infer from the job titles and descriptions where each competitor is investing — sales, product, enterprise, international — and what that signals about their roadmap.

5 Ask Claude to write a comparison table across all competitors using everything you've pasted in: positioning statement, pricing tier structure, top customer complaints, apparent growth vector, and one-line differentiation from your product.

6 Take that comparison table into a second Claude session with your own product's positioning pasted in, and ask it to identify the whitespace — what pain points competitors aren't addressing and where your messaging could be sharper.

7 Drop the final synthesis into a doc and share it. Then repeat this manually next quarter, or next month if something significant shifts.

Prompts you can copy

Here are the homepage and pricing page text for three competitors: [paste]. Extract each company's core value proposition, who they're targeting, and what pricing model they're using. Format as a comparison table.

Here are 20 G2 reviews for [Competitor A] and 20 for [Competitor B]: [paste]. Summarize the top three recurring complaints and top three praised features for each. Note any patterns that suggest a gap a third product could fill.

Here are open job listings from [Competitor A]'s careers page: [paste]. Based on these roles, what are the most likely areas they're investing in over the next 12 months? What does this suggest about their product roadmap or go-to-market focus?

I run a [brief description of your product] targeting [ICP]. Here is a competitive landscape summary I built: [paste]. Where is the clearest whitespace in positioning? What objections does my product currently leave unanswered that competitors are addressing?

Summarize recent news, product launches, and funding announcements for [Competitor Name] from the past 90 days. Cite your sources.

Reality check

Where this gets hard

The walkthrough above works — until your numbers change, the LLM hallucinates, or you have to re-paste everything next month.

No memory between sessions — every research run means re-pasting all competitor content from scratch, even if you did this exact workflow three weeks ago.

Public web data goes stale fast. Competitor pricing pages, job boards, and positioning change frequently, and there's no alert when something important updates.

Pasting raw HTML from competitor pages is messy. You get navigation text, footer links, and cookie banners mixed in with the actual content, which degrades summary quality unless you clean it first.

The comparison table structure you carefully prompted this month won't match what you get next month. Outputs drift, so maintaining a consistent competitive record across time requires manual reformatting every run.

No connection to your own business signals — customer win/loss data in HubSpot, inbound mentions on X, or traffic referral patterns in PostHog. The LLM only sees what you paste, not what your business is already measuring.

Nothing is shareable in a living format. You get a text output that you copy into a doc, and that doc starts aging the moment you close the chat.

Tired of the friction?

Starch runs the whole workflow on live data — no copy-paste, no hallucinated numbers, no re-prompting next month.

See the Starch version →

Starch alternative

The same workflow on Starch

Starch is an agentic operating system — an agent builds and runs the actual software your workflow depends on, continuously, against live data. For competitive research, that means a persistent system that tracks competitors, monitors your own signals, and surfaces changes automatically rather than a one-off prompt you re-run by hand.

Track brand mentions on X continuously using the X Mentions Tracker starter app — Starch automates daily monitoring through your browser and logs every mention, so you catch competitor comparisons and customer sentiment shifts as they happen, not in a quarterly review.

Connect PostHog through Starch's integration catalog and use the Growth Analyst app to surface referral traffic and conversion trends by channel on a weekly schedule — so when a competitor's marketing move changes your traffic mix, you see it in the digest, not months later.

Describe the competitive tracker you want in plain English — 'Build me an app that stores competitor positioning, pricing, and top G2 complaints for five companies, and lets me update notes each quarter' — and Starch builds it as a persistent app, not a chat session you lose.

Starch automates competitor career pages and pricing pages through your browser — no API needed — so you can schedule a weekly pull of open roles and flag new postings without manually checking five different job boards.

Set up an automation: 'Every Friday, pull my X mentions from this week, check for any references to [competitor names], and Slack me a summary' — Starch runs it on schedule so competitive signal collection stops requiring your attention.

Once your competitive data is structured inside Starch, connect it to your HubSpot deals from the integration catalog — so you can ask 'which competitors show up most in deals we've lost this quarter' against live CRM data, not a static export.

Get closed-beta access →

Toolkit