Last updated: October 29, 2026
46% of US business decision-makers say they trust review sites as an information source, and 39% say the same about AI chatbots. That comes from a survey commissioned by a social platform and run on a survey company's panel, which asked 1,202 people between December 23, 2025 and January 7, 2026. Peer recommendations topped the trust list at 73%. Social media came last at 36%.
My verdict: treat your G2 and Capterra profiles as fact sheets that AI assistants will paraphrase, and spend your effort on two things. Keep every field accurate and current, and collect a steady stream of honest reviews through each platform's own tools. Skip anything that smells like review manipulation. In 2026 the platforms' rules and the FTC both punish it, and no ranking gain is worth that.
Those numbers matter for AI search. When a buyer asks an assistant for a shortlist, the answer often draws on review platforms, and that same buyer trusts the review sites a little more than the assistant summarizing them. So your profile shapes the answer and also the page the buyer clicks to check it.
What the survey can and can't tell you
Reddit commissioned the study, called "The hidden B2B journey," and SurveyMonkey ran it on its panel. The two companies published it on March 10, 2026, with a modeled error estimate of plus or minus 3 points. Keep two caveats in mind before you quote it in a deck.
First, Reddit sells advertising to B2B marketers, and the press release makes the case for Reddit as a research channel. That doesn't make the trust figures wrong, but the sponsor had a reason to want peer channels to look strong.
Second, the respondents were decision-makers for every kind of business purchase. Software buyers were one group in the sample. The question also measured trust in each source. It says nothing about which channel buyers open first or how often they use it. Read the 7-point gap between review sites and chatbots as a sign of relative trust and leave it there.
The scope of this guide is narrow on purpose: G2 and Capterra profiles. For how answer engines pick sources in general, start with the AEO pillar guide. If ChatGPT is the assistant you care about most, the guide on how to get recommended by ChatGPT goes further.
Why assistants lean on review profiles
A review profile packs a lot into one URL. It has a product description, a category, feature lists, pricing notes, a rating and often hundreds of first-hand comments. The layout is the same for thousands of products, so it's easy for any system to line one vendor up against another.
Analysts see the same pattern. In a blog post dated June 8, 2026, Forrester analysts Amy Bills and Karen Tran wrote that review content from G2, TrustRadius, PeerSpot and Reddit "substantially informs" the models behind ChatGPT, Claude, Perplexity and Copilot, and that marketers "must include review-site presence" in their AEO plans. That's analyst opinion with no measurement behind it in the post, though it matches what you'll find if you run your own category prompts and read the citations.
Assistants rarely just link your profile. They restate it in their own words. If your G2 description still says you serve "small teams" two years after you moved upmarket, expect that phrase to show up in an AI summary of your company.
Get the category right before anything else
Your category decides which comparisons you show up in, both on the platform and in AI answers built from it.
On G2, the category shown at the top of your profile is your main category. G2's research FAQ says it's also called the "primary" category. According to G2's product information documentation, last updated August 26, 2026, you can switch it with the Change Main Category dropdown, but only to a category already associated with your product. To change the full list, you use Request Updates to reach G2 Support. Each category page has a "What is [category name]?" section, and Show More opens its inclusion criteria, so check those before you ask.
If nothing fits, G2 accepts requests for new categories, which its Market Research team reviews. The request has to list at least 10 products that would belong there. Think twice before you file one. A new category with a handful of products in it won't be the comparison buyers run, and a strong position in an established category is usually worth more.
Capterra gives you less control. Its Profile Guidelines, last updated May 4, 2026, say a product "must fit, at our sole discretion, within one of the existing categories." Your best lever there is an accurate description and real screenshots.
Try this before you change anything. Write down the three prompts a buyer is most likely to type to find a product like yours, for example "best [category] software for [industry]." Then find the G2 and Capterra category pages those prompts would map to. If you're missing from the most important one, fix that first.
Fill in the fields assistants quote
Think of the profile as a structured fact sheet that a buyer and a model will both read. Ad copy works against you here.
The product description
G2's documentation says the "What is [your product name]?" field takes up to 500 words. You edit it in my.G2 under Product Profile, then Product Information. Write it the way you'd want an assistant to describe you:
Open with what the product is and who it's for: "[Product] is [category] software for [audience] that [main job]."
Follow with the three or four capabilities buyers compare on, in plain words.
List the integrations and deployment options people filter by.
Name the company sizes and industries you serve, if that's true and stable.
G2 also has an Optimize for SEO button that generates AI-enhanced text from what you typed. If you use it, read the output line by line. Generated copy drifts toward vague claims, and vague claims are the ones you least want an assistant repeating.
Capterra is stricter on tone. Its Profile Guidelines say not to write in the first person, not to use "overly comparative language and superlatives (e.g., best, most, fastest)" and not to put calls to action in the description. They also say the final copy "is subject to the discretion of our content team." If you write something their editors won't need to rewrite, your wording has a better chance of surviving.
Screenshots and video
Capterra's guidelines require screenshots that show "the software's real user interface" and ban stock images and marketing jargon in them. Real screens help buyers too. A hero image with a slogan tells nobody what the product does.
Pricing and plans
Where a platform lets you add pricing, keep it accurate and log the date you last changed it. A wrong starting price is an easy fact for an assistant to pick up and repeat, and a buyer who sees one number in an AI answer and another on your site has a reason to doubt both. If you only sell on quote, say that plainly in the field.
G2's AI tags
G2 offers two AI tags, AI Enabled and AI Verified, and a product can carry only one at a time. You can turn on AI Enabled yourself in my.G2. G2 upgrades it to AI Verified once 10 or more approved reviews confirm customers have used the AI features. If your product has real AI functionality, turning the tag on takes a minute. If it doesn't, leave it off, because G2's Market Research team can change the tag based on the evidence it finds.
Collect reviews the way the platforms allow
A steady flow of honest reviews beats a perfect average built from one launch push. G2's community guidelines don't accept reviews based on experiences more than two years old, so a profile that stopped collecting in 2024 is already losing usable material.
Here are the rules as each platform states them, checked on October 21, 2026.
G2's community guidelines cap the value of any review incentive at $100, and the cap covers cash, gift cards and non-cash items like swag, credits and subscriptions. Incentivized reviews get a label. G2 also bans what it calls biased review collection, which includes any method that "segments out potentially negative reviews or exclusively targets positive reviews," and incentives that depend on a positive rating.
Capterra's Community Guidelines, also last updated May 4, 2026, apply to every vendor. You must ask "a broad cross-section" of your users and can't target people based on expected sentiment. You can't coach customers on what to write, and you can't use third-party services that churn out large volumes of unverified reviews. Vendors in its Reviews Collection Program have extra rules:
Incentives go equally to every eligible participant, whatever rating they give.
You have to use the incentive link from the Vendor Portal.
Every outreach message must say the incentive doesn't depend on a positive review.
Capterra also addresses non-disparagement clauses. Its guidelines say federal law prohibits them in customer contracts and that Capterra won't remove a review because the reviewer signed one.
The FTC has shown it will enforce this. On July 15, 2026, it finalized an order against TruHeight, a supplement company, and its two principals. The order carries a $4 million judgment, suspended after a $750,000 payment because they couldn't pay the full amount. According to the April 13, 2026 announcement of the complaint, employees wrote thousands of the five-star reviews on TruHeight's website until at least November 2024, and the company offered free or discounted products for five-star reviews on its own site and on third-party platforms. The complaint charged violations of the FTC Act and the agency's Reviews and Testimonials Rule. It's a consumer case in another industry, and the conduct ran from 2021 to late 2024, but that rule isn't written for one sector.
A cadence that holds up looks like this. Ask customers when they hit a clear milestone, such as finishing onboarding or a first renewal. Ask everyone in that segment, including the accounts your CSMs worry about. Send people through each platform's own review link. Spread requests across the year so the profile never goes quiet for a full quarter.
Reply to criticism and to reviews that describe an old product
Expect some negative reviews. The ones that hurt in AI search are the unanswered complaints and the old reviews about features you've since fixed.
G2's guidelines give sellers "the opportunity to respond publicly" to discussions on their profile, and invite them to give their side when a post is inaccurate. Capterra says replies have to go through its own tools, such as the Vendor Portal reply feature. In each reply:
Acknowledge the specific issue in one sentence.
Say what changed and when: "Since [month year], [feature] supports [capability]."
Link the help article or release note that documents it.
The dated second step does the most work. An assistant summarizing your reviews may pick up a complaint from 2024, and a specific, dated reply on the same page gives it the correction right next to the problem.
Don't argue in replies, and never ask a reviewer to change a rating. Capterra's guidelines forbid pressuring reviewers to remove or modify honest reviews, and G2's forbid contacting reviewers to influence or coerce them.
If the wrong information is appearing in AI answers and the profile itself is fine, you need a wider fix. The guide on fixing wrong information ChatGPT gives about your company covers that process.
Where to start, based on your situation
If you've never claimed your profiles, claim both, write the description using the structure above, upload real screenshots and set the correct main category. For most teams that's a few days of work.
If you have profiles but few reviews from the past year, set up a milestone-based request through each platform's official link and aim for a small number of reviews every month.
If you've moved upmarket or repositioned, rewrite both descriptions first. Then reply to the top-voted reviews that describe the old product, with dated corrections.
If you're in the wrong G2 category, check whether the right one is already associated with your product and switch the main category. If it isn't, file Request Updates with evidence that you meet the inclusion criteria.
If you're a newer product with almost no reviews, don't chase a big number. Get your first 10 honest reviews from active users, make the description precise and treat category placement as the main job for your first quarter.
How to tell whether it's working
Pick a fixed set of category prompts and run them every month in the assistants your buyers use. Note which review-site pages get cited and how your product is described, then compare that wording with your profile. The walkthrough on measuring AI search visibility covers how to keep prompt wording from skewing your results, and the generative engine optimization primer explains why assistants paraphrase sources the way they do.
Watch referral traffic too. G2's documentation says it appends a UTM Content value to your website links that identifies which G2 page sent the click, and the Profile Activity tab in my.G2 tracks clicks to your site. If review-site work is one part of a wider AI visibility program, Miraa's answer engine optimization page describes how those pieces fit together.
What to watch out for
Sending review requests only to customers with high satisfaction scores is review gating. It breaks both platforms' rules, and it's the kind of practice the FTC's rule targets, so ask the whole segment.
Reviews from your own staff or partners are off limits. G2 bars employees of the vendor and of direct competitors, and Capterra also bars anyone affiliated with the vendor or with a financial interest in the product.
Superlatives get edited out on Capterra and read as marketing everywhere else. Specific, checkable claims serve buyers and assistants better.
It's easy to keep G2 tidy and let Capterra go stale, or the other way around. Assistants pull from both.
Your product changes every quarter, so a profile is never finished. Put a profile review on the same calendar as your pricing page review.
Final thoughts
A review profile is one of the few third-party pages about your product that you can partly shape, which makes it worth real effort and makes shortcuts tempting. The work that pays off is plain: the right category, a description an assistant could quote without embarrassing you, reviews from a broad set of customers through official links, and dated answers to criticism. Repeat it every quarter, because your product and your reviews keep changing.