Software Review Sites Sold for $110 Million, After a $150 Million Write-Off

Gartner sold three software review sites to G2 on 5 February 2026 for about $110 million. One quarter earlier it wrote $150.0 million of goodwill off the same business. The seller destroyed more value than the buyer paid, and that gap is the clearest public price anyone has put on the category. Both figures come from Gartner's own filings.

The short answer

Software review sites are not dead, but their business model is. Traffic that funded lead generation collapsed, while their reviews became the raw material AI assistants cite for free. Gartner's exit at $110 million prices the transition honestly. Expect the category to survive as a data supplier, not a destination.

What the $110 million actually says

Gartner spent a decade assembling Digital Markets. Software Advice came in 2014. Capterra and GetApp followed. Together they became one of the largest software discovery businesses in the world.

On 29 January 2026, G2 announced it would buy all three. The announcement quoted scale, not price: about 6 million verified reviews, more than 200 million annual software buyers, 10,000 vendors and 2,000 categories. No financial terms were disclosed.

The price only surfaced in the filings. Gartner's Form 10-K states that on 5 February 2026 it completed the sale for approximately $110.0 million, before customary adjustments. That is the whole of a business assembled over eleven years.

Read the two filings together and the story sharpens. Gartner's Form 10-Q for the September 2025 quarter records a $150.0 million goodwill impairment in the Digital Markets reporting unit, triggered because "ongoing weakness in the market" forced a revision to the unit's long-term earnings forecast.

The same filing quietly demoted the business. Digital Markets stopped being part of the Insights segment and moved into "Other", on the basis that it no longer met the criteria of a reportable segment. Companies do not usually hide a business they intend to keep.

Gartner does not disclose Digital Markets revenue separately, so there is no public multiple to compute here. The absence of that line is itself part of the record.

The traffic went first, and it went everywhere

A price is a conclusion. The cause sits upstream, in search.

SE Ranking measured organic search traffic to 23 review platforms and found declines that are hard to describe as a downturn. G2 fell about 84.5%, from roughly 2.56 million monthly visits to about 397,000. Capterra fell about 89%, from roughly 1.63 million to about 179,000.

The rest of the field looks the same. TrustRadius was down 92.2%, Software Advice 86.5%, and Gartner Peer Insights 76.5%. When every platform in a category moves together, the cause is the channel, not the operator.

Read the methodology before you quote the number

These figures are estimates, and I would not treat them as audited. Similarweb models visits rather than counting them, and the study is a single snapshot taken on 1 December 2025 across 30,000 keywords.

They are still the best public series available, because none of these platforms publishes traffic. Treat the direction as solid and the decimal places as decoration.

Citations went up while visits went down

The same study found something more interesting than the decline. Review platforms are among the most cited sources inside Google's AI Overviews, and five platforms account for 88% of all review platform citations.

Gartner Peer Insights took 26.0% of those citations. G2 took 23.1%. Both lost the overwhelming majority of their visitors anyway.

That is the whole disintermediation story in one sentence. The content is still authoritative, and the page it lives on is no longer visited. The same split is showing up across categories, and the underlying mechanics are set out in the analysis of how citation and rank came apart.

The trust problem is older than the AI problem

It is comfortable to blame ChatGPT for this. That story is too clean, and it lets the category off a hook it earned.

Buyers had been discounting review site rankings for years, because the ranking and the advertising were sold to the same people. Sponsored placement, paid profiles and vendor-funded review drives are all disclosed practices. Disclosed is not the same as trusted.

What the FTC rule bans, and what it allows

In August 2024 the FTC finalised a rule banning fake reviews and testimonials, effective 21 October 2024. It prohibits buying, selling or creating fake reviews, including reviews written by anyone with no experience of the product and reviews generated by AI.

On incentives the rule is narrower than most people assume. It bans incentives that are conditional on sentiment, expressly or by implication. Paying every reviewer the same amount regardless of what they write remains legal.

The FTC can seek civil penalties against knowing violators, set at up to $51,744 per violation at the time of announcement. That ceiling is adjusted annually, so check the current figure before quoting it.

Enforcement reached the platform layer within months. In January 2025 the FTC approved a final order against the review platform Sitejabber, which collected ratings at the point of sale, before buyers had received what they bought, and folded them into displayed averages. The order runs for 20 years.

The gift card is not the real problem

G2 runs vendor-funded review campaigns with gift card incentives, labels the resulting reviews, and forbids conditioning the reward on a positive rating. That is a defensible design, and it is more transparent than most of the market.

The distortion is upstream of the individual review. A vendor with budget runs a campaign and collects 200 reviews in a quarter. A better product with no campaign collects nine. The star ratings are honest and the ranking still measures marketing spend.

My position is that this was always the weak point, and AI assistants simply made it legible. When a buyer asks a model to compare three tools, the model does not care which vendor bought placement in a grid. Vendors trying to influence that layer instead should read the working notes on how shortlists now form inside an assistant.

The business was leads, not opinions

Review sites present as publishers. They are priced and operated as lead generation businesses.

Reviews are the acquisition cost. Rankings are the retention mechanism. The revenue arrives when a vendor pays for placement, for a profile, or for a buyer's contact details.

Why a lead generation model breaks first

A media business with falling traffic loses advertising revenue in proportion. A lead generation business loses it faster, because leads are the only unit sold and there is no brand premium to fall back on.

An 84.5% traffic decline is not an 84.5% revenue decline, since paid placement and intent data still monetise a smaller audience. It is enough to force a forecast revision, which is exactly what Gartner disclosed.

Pay-per-lead is the bet G2 just doubled

G2's acquisition announcement is explicit about the plan. It promises three times the buyer intent signals, expanded AI-driven recommendations through G2.ai, and a new global pay-per-lead offering.

That is a coherent response. If buyers no longer visit, sell the vendor the signal that a buyer looked. It also concedes the point: the destination is worth less than the data it generates.

G2's own research argues both sides

This is where a careful reader should slow down, because the best available buyer data comes from a party to the transaction.

In April 2026, G2 published The Answer Economy, based on 1,076 buyers surveyed in March 2026. It found 51% start research in an AI chatbot more than in Google, up from 29% in April 2025. It named AI chatbots the top influence on which vendors make a shortlist. It also found 45% still rate review sites as the most confidence-inspiring signal, and that 64% encounter inaccurate AI recommendations often or very often.

Three months later, G2 published The Evaluation Maze, based on 1,038 decision-makers surveyed in June 2026. In that survey review sites led shortlist influence at 38%, with AI chatbots second at 37%.

Different samples, different question wording and a one point gap in the later result. This is not evidence of a reversal, and it is not evidence of a rout. It is evidence that the honest answer sits inside the margin of error.

I would not accuse G2 of cooking either number. Both were published, both carry a stated sample, and the March survey contained findings that damage its own franchise. That is more disclosure than most vendors offer, and it is the standard argued for in the case for publishing original research with its methodology attached.

What I would not do is build a budget on a one point difference between two surveys with different samples. The defensible reading is that AI assistants and review sites are now roughly equal inputs, and that neither is dominant.

Where this argument is weakest

The case above has three genuine holes, and I would rather name them than have a reader find them.

The case that this is a good acquisition

A distressed price is not the same as a dying asset. Gartner is a research and advisory business whose CEO said plainly that Digital Markets did not fit the strategy. A conglomerate shedding a business it never operated well tells you about the seller.

G2 is a specialist buying assets adjacent to its own. It removes its three closest competitors, consolidates roughly 6 million reviews, and gains the intent data behind them. Bought at $110 million, that could work well even if category traffic never recovers.

The part nobody can settle

AI assistants need a source of structured, verified opinion about software. Nobody has built a better one than a review corpus with named reviewers and verification steps.

So there are two futures, and the public evidence does not choose between them. In one, review sites become an unpaid substrate whose value is captured by the models that cite them. In the other, they become a licensed data supplier and get paid per query. The licensing deals that would settle it are mostly private, and the measurement problem underneath is covered in the review of what the citation evidence actually supports.

Anyone telling you confidently which future arrives is guessing. The indicator I watch is whether any review platform announces a paid content licence with a model provider, because that single event would reprice the category.

What survives, and what stopped working

Strip the narrative away and the practical question is narrow. Which parts of a review site budget still do work?

Three things survive. Verified reviews still function as proof for a buyer already considering you. Category presence still matters for retrieval, since assistants pull from pages that exist. Intent data still identifies accounts researching your category, and that signal is independent of whether anyone clicks.

Three things stopped working. Buying placement to win a category page does little when the page has lost most of its visitors. Badge collection produces artwork rather than pipeline. Review volume as a vanity race is now the most expensive way to buy a rank nobody sees.

The last row is the substitution most teams have already made informally. The economics of doing it deliberately are set out in the template for B2B comparison pages.

One more discipline is worth stating. When you cut a channel, you need something that proves the replacement worked, and reviews are weak evidence compared with a customer outcome you can verify. The bar for that is set in the standard for case studies with verified outcomes.

Frequently asked questions

How to reset the review site budget this quarter

Two moves, both possible before your next renewal conversation.

First, ask your platform account manager for monthly sessions on the specific category pages you sponsor, going back 24 months. Not impressions, not audience reach, sessions. If they cannot or will not produce it, you have learned the price of the page without needing the number.

Second, take one product and ask ChatGPT, Claude and Perplexity to recommend tools in your category, three times each with different phrasing. Write down which sources each model cites. That list, not a grid position, tells you where the next quarter of budget belongs, and the measurement approach behind it is set out in the guide to measuring discovery you cannot see.

Related reading