The 50 fastest-growing SaaS companies barely exist in AI answers

Why this question matters, and to whom

A company that ranks first for its category on Google has spent years, and often a great deal of money, winning a very specific game: being the blue link a searcher clicks. That game has a scoreboard everyone agrees on, ranking position, and a large industry exists to help companies win it.

AI assistants are a different game with, so far, no public scoreboard. When a buyer asks ChatGPT or Google AI Overview "what's the best tool for X," the assistant does not show a ranked list of ten blue links. It picks a small number of names, sometimes one, and says them in prose. If your company is not one of the names picked, you are not clicked, not considered, and in a growing number of buying journeys, not known to exist for that search at all.

This matters most acutely to companies that already believe they have won the visibility problem because they can point to organic search growth. The fifty companies in this cohort are, by Ahrefs' published methodology, the ones who grew fastest by that measure. If classic search dominance guaranteed AI-answer visibility, this is the cohort where we would expect to see it most clearly. It matters to marketing and growth teams deciding where to spend the next budget cycle, and to anyone advising a SaaS company on whether their SEO investment is future-proof against how buyers are starting to search.

It also matters because the two forms of visibility are not obviously substitutable. A company invisible to AI answers does not get a second chance the way it might in classic search, where a buyer scrolls past the first result to the second or third. Assistants tend to name very few companies per answer, so absence from the answer is closer to absence from the market for that specific query than it is to ranking eleventh on a results page.

How the measurement works, and why it was designed this way

The unit of analysis is one row per (company, prompt, engine). Each of the 50 companies in the Ahrefs cohort was assigned a category by the study team, for example, this is an expense-management tool, or this is an AI video generator, and then asked about using prompts phrased the way a real buyer would type them, generic category questions rather than questions that name the company. This produced 200 distinct prompts run across 4 engines: ChatGPT, Claude, Gemini, and Google AI Overview, for a total of 800 rows.

The design choice worth explaining is that every company is judged only on the question its own buyers would ask. A payroll company is asked about payroll software, not about video generation. This avoids a common failure mode in AI-visibility studies, where a company is marked invisible simply because it was asked about the wrong thing. It also means each company's result is comparable to the others in kind, if not in raw difficulty, since categories vary in how crowded they are.

Each engine was run once per prompt per company, with model versions pinned: 1 version of ChatGPT (gpt-5.5-2026-04-23), 1 version of Claude (claude-sonnet-5), 1 version of Gemini (gemini-2.5-pro), and Google AI Overview as it rendered live. The run happened over a window of 2 timestamps in August 2026, not a single instant, which matters because these systems are not static from one day to the next and a single-timestamp study risks capturing a transient state rather than a durable one.

Of 800 rows, 773 returned a usable answer, 8 rows were cases where Google did not show an AI Overview at all for that prompt, and 19 could not be parsed into a clean mention or non-mention. The 773 usable answers are the basis for every percentage in this study. Readers should treat the 19 unparsed rows as a small source of noise, not as hidden non-mentions or hidden mentions.

The headline result, and what it does and does not mean

Across 773 usable generic-prompt answers, the cohort was mentioned 66% of the time. Read alone, that number looks respectable, roughly two answers in three include one of these fast-growing companies. The average, however, is the least informative number in this dataset, because it blends two very different populations of companies into one figure.

What the headline number does show is that being named by Ahrefs as a fast-growth SaaS company does not, by itself, predict being named by an AI assistant. If it did, we would expect a rate much closer to universal, since every company in this cohort has already demonstrated it can win a discovery channel that AI assistants partly draw on, organic content and reputation. Instead, a meaningful share of the cohort is structurally absent, a result covered in the next section.

What the headline number does not show, and cannot show, is why any individual company is missing. Absence from an AI answer is consistent with several different stories: the company may be too new or too narrowly positioned for the category framing used here, the assistant's training or retrieval data may under-represent it relative to more established incumbents, or the category itself may be dominated by a small number of brand names the assistant defaults to regardless of who is actually best suited to a buyer's need. Distinguishing between these would require testing the same companies against more specific, more branded, or longer-tail prompts, which is outside what this dataset captures.

The study also cannot show that the invisible companies are inferior products or worse businesses. Ahrefs selected this cohort for organic search growth, not for AI-answer readiness, and the two are plausibly measuring different things: one measures whether a company's content and backlink profile grew quickly, the other measures whether a language model's internal sense of a category includes this name at all. A company can be growing fast and still be new enough, or narrow enough, that it has not yet accumulated the kind of broad, repeated, third-party mention density that seems to be what gets a company named.

Is the effect gradual or binary, and what follows from that

The shape of a result changes what a company should do about it. A gradual effect, where every company gets a little visibility and the fast-growing ones get somewhat more, implies a tuning problem, spend a bit more effort and move up gradually. A binary effect, where companies split cleanly into visible and invisible, implies a threshold problem, something has to change qualitatively before a company crosses over at all.

This dataset is closer to binary. 5 of the 50 companies, 10% of the cohort, received zero mentions on every one of the 4 engines tested: Cartpanda, ClipDrop, DocuClipper, Fyle, and Tavus. These are not companies that scored low, they scored nothing, on every engine, across every prompt asked about their category. At the other pole, a set of companies, Vercel, Supabase, Cradlewise, Rocketlane, and Storylane, were the most consistently visible in the cohort. In between, 10 companies, 20% of the cohort, were visible on some engines and not others, which is itself a form of binary outcome repeated per engine rather than a smooth in-between state.

The practical implication is that a company on the wrong side of this line should not expect steady, linear improvement from doing a bit more of what it is already doing. If the zero-visibility group is missing because assistants default to a small set of incumbent brand names for that category, as the displacement data suggests, then incremental content production is unlikely to move the needle. What would move it is a change in whether the assistant's model of the category includes this name at all, which is a different kind of work, closer to earning mentions on the specific sources each engine reads, than to producing more general marketing content.

The caveat is that ten of fifty and five of fifty are both small enough that individual companies moving between groups on a re-run would shift these percentages meaningfully. This study ran each prompt once per engine within the 2-timestamp window, so it captures a single measured state, not a distribution across repeated runs.

What the engines are reading when they decide who to recommend

Each engine cites a different mix of sources, and the differences are large enough to matter to anyone deciding where to invest.

Across all engines combined, 30 distinct third-party sources appeared often enough to register as a top citation, with review platforms (g2.com, learn.g2.com, capterra.com), analyst sites (gartner.com), and, notably, informal sources (reddit.com, youtube.com) all appearing among the most-cited. ChatGPT was recorded citing 12 distinct top sources, Claude 12, and Google AI Overview 12, while Gemini's citation behavior was concentrated enough that only 1 distinct top source, google.com, was recorded, which likely reflects Gemini surfacing its own search results as the visible citation layer rather than reading a comparably diverse source set.

The pattern worth noting is that none of the four engines lean primarily on a company's own website or its organic content, the very asset that won these companies their Ahrefs ranking. Instead, three of the four engines lean heavily on third-party review sites, G2 and Capterra chief among them, plus analyst content from Gartner. Google AI Overview and Claude both draw meaningfully on informal, user-generated sources, Reddit and YouTube, sources a company cannot directly author but can be discussed on.

This has a direct, actionable reading for a marketing team. Ranking first in Google's organic results measures something these engines are not primarily reading. A company's presence, review volume, and framing on G2, Capterra, and Gartner, and its discussion volume on Reddit and YouTube, appear to matter more to whether an assistant names it than the strength of its own owned content. That is a different budget line than most SEO programs currently fund.

Where the engines disagree, and why one visibility score is misleading

Vendors selling a single "AI visibility score" implicitly assume the four engines broadly agree on who deserves to be mentioned. This dataset does not support that assumption.

Appearance rates alone vary by a wide margin: Google AI Overview mentioned cohort companies in 82% of its answers, ChatGPT and Claude both at 80% and 80% respectively, and Gemini lowest at 74%. A company sitting near the top of Gemini's threshold for a given category is not guaranteed a mention on Google AI Overview's threshold for the same category, and the 10 companies visible on some engines but not others are the direct evidence of that: these are real companies where the same prompt, asked of the same underlying business, produced a mention on one engine and silence on another.

Position, for the answers where a mention did occur, also differs by engine. The average position at which a cohort company was mentioned, when mentioned, was 2.48 on Google AI Overview, 3.07 on ChatGPT, 3.24 on Gemini, and 3.66 on Claude, against an overall average of 3.11 across 510 scored mentions. Google AI Overview not only mentions cohort companies more often, it tends to mention them earlier in the answer when it does.

The source tables in the previous section explain part of this: an engine that reads Reddit and YouTube heavily is answering a different underlying question, in effect, than an engine that reads Microsoft and AWS documentation heavily, even when given the identical prompt. A company well-discussed on Reddit but thin on enterprise documentation platforms should expect to do better on Google AI Overview or Claude than on ChatGPT, and the reverse for a company with deep technical documentation but little community discussion.

The practical consequence is that a single blended score, averaged across engines, would report a company as "moderately visible" when the true picture is "highly visible on one engine, invisible on another," which are different problems requiring different fixes. A buyer of AI-visibility tooling should ask for the per-engine breakdown, not the blend.

What a company on the wrong side of this should actually do

For the 5 companies with zero visibility on any engine, the displacement data gives a specific, useful clue. When Cartpanda is not mentioned, the assistants name Shopify, BigCommerce, Stripe, Shop Pay, and WooCommerce instead. When ClipDrop is not mentioned, they name Adobe Photoshop, Luminar Neo, Canva, Adobe Firefly, and Topaz Photo AI. When DocuClipper is not mentioned, they name Docparser, Lido, Nanonets, Parseur, and Google Document AI. When Fyle is not mentioned, they name Ramp, SAP Concur, Navan, Expensify, and Brex. When Tavus is not mentioned, they name Runway, Google Veo, HeyGen, Kling AI, and Synthesia.

Each of the 5-company displacement lists shown for each of the five zero-visibility companies is dominated by large, long-established incumbents, not by other fast-growing peers from the same Ahrefs cohort. That pattern is worth sitting with: it suggests these engines default to brand-name incumbents for a category unless given a specific reason to reach further down, rather than defaulting to whichever company has grown fastest recently.

The corresponding action is not to write more blog posts. The source data in the earlier section shows these engines read G2, Capterra, Gartner, Reddit, and YouTube more than they read a company's own site. A company in a zero-visibility position should first check whether it has any meaningful presence on those specific platforms, a G2 or Capterra profile with real review volume, a Gartner listing, discussion threads on Reddit that name the product in the context of the category, and comparison or demo content on YouTube. If those are thin or absent, that is a plausible, testable explanation for the silence, and it is fixable in a way that more SEO content is not.

A second, harder-to-fix possibility is that the category itself is one where a handful of incumbents have accumulated so much mention density that a newer entrant needs a sharper, narrower category claim, competing to be named for a sub-category the incumbents do not own, rather than the broad category where they dominate. A company should test both its current broad category prompt and a narrower variant of it before concluding the fix is purely a citation-building problem.

How to read the published dataset yourself

The dataset ships at the row level, one row per (company, prompt, engine), so a reader can re-aggregate it however they choose rather than trusting only the cuts in this report.

The most important caveat to carry into any re-analysis is the category_source=assigned flag on every row. The category tested against each company was chosen by the study team, not published by Ahrefs, and it is the single most consequential methodological choice in the study. Ask about the wrong category and a real, healthy company will look invisible for reasons that have nothing to do with its AI-answer readiness. The published category column lets a reader substitute their own label for any company and see whether the result holds.

A few counts to keep in view when re-slicing. Of 800 total rows, 773 produced a usable, parseable answer. 8 rows are cases where Google simply did not render an AI Overview for that prompt, which is a Google product decision, not a company absence, and should not be counted as a non-mention. 19 rows could not be reliably parsed and should be treated as missing data, not as evidence either way. Any re-aggregation that silently drops these distinctions will produce a different, and less defensible, appearance rate than the 66% reported here.

Engine version matters too. Each engine was pinned to a single version for the entire run, 1 version of ChatGPT, 1 of Claude, 1 of Gemini, and Google AI Overview as it rendered live across the 2-timestamp window. A re-run months later against newer model versions is not guaranteed to reproduce these figures, and readers should treat this as a snapshot of a specific moment in a fast-moving set of products, not a permanent characterization of any engine's behavior.