How to Automatically Extract Themes from G2 Reviews

Quick answer: to extract themes from G2 reviews automatically, put the reviews in a spreadsheet with one review per row, keep G2's "What do you like best?" and "What do you dislike?" answers in separate columns, and run each column through an AI thematic analysis tool. You get a short list of named themes for strengths and for complaints, each with a count, a percentage and example reviews.

G2 reviews are unusually useful for this because the platform asks every reviewer the same structured questions. That structure does half of the analysis work for you, if you keep it intact.

Why G2 reviews are worth analyzing

• They are written by buyers. Reviewers describe the product in the words your prospects use, which is useful for positioning and copy.
• They are structured. Separate answers for likes, dislikes and problems solved mean you can analyze praise and criticism without mixing them.
• Competitors have them too. The same method works on a competitor's reviews, which shows you their recurring weaknesses.

The problem is volume. A product with 300 reviews has around 900 separate answers. Reading them all is a day of work, and reading a sample gives you anecdotes rather than numbers.

Step 1: Collect the reviews you are allowed to use

Start with reviews of your own product. If you manage your product's G2 profile, check your G2 vendor account for export or integration options. Otherwise, copy reviews into a spreadsheet by hand. For competitor reviews, read G2's terms of use before collecting anything, and avoid automated scraping where it is not allowed.

Create one row per review with these columns:

Column	Example
Review ID	G2-0142
Date	2026-08-19
Star rating	4
Reviewer segment	Mid-market, Marketing
Likes	"Setup took ten minutes and the templates are great"
Dislikes	"Reporting is basic and exports are clunky"
Problems solved	"We stopped tracking campaigns in five spreadsheets"

Step 2: Keep likes and dislikes separate

This is the most important step. If you merge everything into one column, every theme mixes praise and complaints ("Reporting" contains both "great dashboards" and "reporting is basic"), and the counts stop meaning anything.

Run separate analyses on:

1. Likes, to find your real strengths
2. Dislikes, to find recurring friction
3. Problems solved, optionally, to learn which jobs customers hire the product for

Step 3: Run AI thematic analysis on each column

In Apercu, the process for each column is:

1. Upload the spreadsheet and select the column, for example Dislikes.
2. Describe the data in the context box: "Answers to 'What do you dislike?' from G2 reviews of a marketing analytics tool."
3. Choose the level of detail. Five to eight themes is usually right for a few hundred reviews.
4. Review the themes, rename any label that is unclear, and move reviews that landed in the wrong theme.

You can try this on up to 500 reviews without an account using the free thematic analysis tool.

Step 4: Read the results

Here is an illustrative example of what the dislikes analysis might return for a mid-sized SaaS product with 280 reviews:

Theme	Reviews	Share
Reporting and export limitations	61	22%
Steep learning curve for advanced features	48	17%
Pricing for small teams	39	14%
Missing or shallow integrations	34	12%
Slow support response	22	8%
Mobile app gaps	17	6%
Other	59	21%

How to read it:

• Rank, do not just list. The top two themes cover almost 40 percent of complaints. That is where a roadmap decision has the most effect.
• Read the examples before acting. "Reporting and export limitations" might be one missing feature or ten small ones. The example reviews tell you which.
• Check the Other bucket. If it is large, rerun with more themes or a clearer context description.

Step 5: Turn themes into decisions

Team	What to do with the themes
Product	Prioritize the top dislike themes and attach the example reviews to roadmap tickets
Marketing	Use the top like themes, in reviewers' own words, in headlines and comparison pages
Sales	Prepare answers for the top dislike themes, since prospects read the same reviews
Customer success	Build onboarding content for learning-curve themes

Analyzing competitor reviews

Run the same likes and dislikes split on a competitor's reviews and compare the two sets of themes side by side. Look for:

• Their top dislike themes that are your strengths, which become positioning angles
• Their top like themes that you lack, which become roadmap input
• Themes that appear for both products, which point to category-wide problems

Repeat every quarter

Save each analysis and repeat it on the latest quarter's reviews. If a fix shipped, its theme should shrink. If a new theme appears, you caught it early.

Common mistakes

• Mixing likes and dislikes in one analysis
• Reading only the newest reviews, which over-weights recent releases
• Treating star ratings as the analysis. A 4-star review often contains the most specific criticism.
• Ignoring reviewer segments. Run separate analyses for small business and enterprise reviewers if their needs differ.

FAQ

Can I export reviews from G2?

If you manage your product's G2 profile, check your G2 vendor account for export or integration options. Otherwise, copy the reviews you need into a spreadsheet, and follow G2's terms of use for any review data you collect.

How many G2 reviews do I need for thematic analysis?

Around 50 reviews start to show clear patterns. With 200 or more, theme percentages become stable enough to compare quarter over quarter.

Can ChatGPT extract themes from G2 reviews?

It can summarize a small batch, but it is unreliable at counting and can paraphrase quotes. A thematic analysis tool assigns every review to a theme, so the counts and examples are traceable. See how to analyze survey responses with ChatGPT for safe prompting tips.

Does this work for Capterra, Trustpilot and app store reviews?

Yes. Capterra has separate Pros and Cons fields that work like G2's likes and dislikes. For Trustpilot and app stores, split reviews by star rating instead. See app review analysis.