What do the two customer-support market scans measure?

Each scan generated 15 real buyer questions for a target company's market, collected the top five ranking public web sources per question, and recorded which companies appeared and how prominently. It measures presence in the sources AI-assisted buyers encounter — not market share, revenue, or product quality.

Which vendors appeared most often in the two July 2026 scans?

Zendesk appeared in 7 of 14 buyer-question source sets in the gorgias.com scan and 7 of 15 in the richpanel.com scan — roughly half in both. Gorgias appeared in 6 of 14 in its own scan and 7 of 15 in the richpanel.com scan, where it tied Zendesk and Intercom for the highest presence.

Why did Richpanel show 0 of 15 in one scan but 4 of 14 in the other?

The two runs asked different question sets. The richpanel.com run's AI-first support questions surfaced no Richpanel mentions (0 of 15), while the gorgias.com run's Shopify-helpdesk questions surfaced Richpanel in 4 of 14 source sets the same morning. Visibility depends on how the buyer frames the market.

What sources do AI-visible answers rely on in this market?

Reddit was the most-cited domain in both scans: 13 of 70 citations in the gorgias.com run and 10 of 75 in the richpanel.com run. Comparison listicles plus community threads made up 41% and 44% of cited sources respectively, and vendor-owned domains such as kustomer.com and fin.ai also ranked as sources.

Can these results be extrapolated to market share or trends?

No. Two single-repetition scans from one day (July 9, 2026) are two data points. Ranked results shift daily, presence is not causation, and this piece deliberately claims nothing about market share, revenue, or trajectory.

The customer-support software market, seen through AI answers

Industry insight · AI Visibility · 8 min read · last verified 2026-07-21

Reviewed before publication Editorial board Independent commercial review

In shortTwo evidence-bound Magrios scans of the customer-support software market, run July 9, 2026, show which vendors appear in the sources AI answers draw from: Zendesk near half of buyer questions, one target absent from all 15 of its own.

Every vendor in customer-support software has a homepage that calls it the leader. AI-assisted buyers increasingly never read those homepages first. They ask a question — what are good alternatives to Zendesk for a growing ecommerce brand? — and an assistant assembles an answer from whatever public sources rank for that question. Whoever is present in those sources gets considered. Whoever is absent does not exist, no matter what the homepage says.

On July 9, 2026, Magrios ran two evidence-bound scans of this market: one centered on gorgias.com, one centered on richpanel.com. Each scan generated real buyer questions for that company's market, collected the top-ranking public web sources for every question, and recorded — with stored evidence spans — which companies appeared, how prominently, and where the answers came from. This piece reports only what those two stored runs contain. Two scans are two data points, not a market census: nothing below is a claim about market share, revenue, or trend lines, and nothing extrapolates beyond the files.

Read together, the runs show a market that changes shape depending on which door the buyer walks in through — and a small set of names, led by Zendesk, that appear no matter which door that is.

Two scans, one day in July 2026

Both runs used the same instrument. For each target company, the engine generated fifteen buyer questions spanning the intents real buyers bring to this category — category discovery, best-solution, direct comparison, alternatives, pricing-and-value, problem-led, use-case-specific, and industry-specific. For every question it collected the five top-ranking public web sources and read three of them in full, extracting each company mentioned along with an evidence span and a confidence score.

The gorgias.com run (started 06:23 UTC) generated 15 questions, of which 14 entered the final analysis, resting on 70 cited sources. The richpanel.com run (started 08:00 UTC) analyzed all 15 of its questions across 75 cited sources. That is 29 question-level snapshots and 145 stored citations of the same market, taken less than two hours apart. Every number below cites one of these two runs.

The gorgias.com scan: visible, but not where deals get decided

In our July 2026 scan of gorgias.com's market, Gorgias appeared in the top-ranking sources for 6 of 14 buyer questions — a 43% presence rate — and was the lead recommendation on 2 of them. Both lead positions were Shopify-shaped: the question about the best helpdesk for a high-volume Shopify store, and the question about support software that surfaces customer order history inside tickets.

Zendesk did better in Gorgias's own market. In the same scan, Zendesk appeared in 7 of 14 question source sets — the highest presence of any company in the run — and took the lead recommendation on 4 questions, double Gorgias's count. Three vendors that were not even on the scan's provided competitor list matched the target's presence: Tidio, Help Scout, and Freshdesk each appeared in 6 of 14 source sets. Behind them, the gorgias.com scan logged Kustomer in 5 of 14, Richpanel and HubSpot Service Hub in 4 each, and Intercom, Gladly, Front, and Zoho Desk in 3 each.

The run's stored findings flag where Gorgias was structurally absent, not merely outranked. In that scan it appeared in none of the 3 problem-led questions (the reduce-resolution-time, automate-refunds genre) and neither of the 2 pricing-and-value questions — the intents closest to a purchase decision. Being present on roundup listicles is one thing; being in the sources when a buyer asks whether an AI agent add-on is worth the money is another. That distinction between showing up and being the load-bearing source is the one we unpack in brand mentions versus citations.

In all, the gorgias.com scan's extraction table named 43 distinct companies and products across just 14 questions, 27 of them appearing exactly once.

The richpanel.com scan: a market that answers without you

Later that same morning, the same instrument scanned the same market from richpanel.com's vantage point and produced the starkest number in either file: in our July 2026 scan of richpanel.com's market, Richpanel appeared in 0 of its own 15 buyer questions.

Meanwhile the incumbents split the room three ways. In that same scan, Gorgias, Zendesk, and Intercom each appeared in 7 of 15 question source sets — a 47% presence rate — and each took the lead recommendation exactly once. Freshdesk and Help Scout followed at 5 of 15; Ada, Sierra, and HubSpot Service Hub at 4 each; Decagon and Tidio at 3; Salesforce Service Cloud, Zoho Desk, and Forethought at 2; Kustomer at 1.

The stored findings sharpen the shutout: Richpanel was absent from all 3 comparison-intent questions and all 3 use-case-specific questions in that run — precisely the high-intent territory where its named competitors were collecting wins.

The long tail is its own story. The richpanel.com scan named 44 distinct companies and products across its 15 questions, 31 of them exactly once. That tail even bleeds outside the category: a single question about the cost of running an AI-powered helpdesk versus traditional per-agent pricing pulled ChatGPT Team, GitHub Copilot, Microsoft Copilot, the OpenAI API, and the Anthropic Claude API into the scan's vendor table, one mention each. When buyers frame support software as an AI problem, AI platforms become part of the competitive set.

And here is the finding you only get from running two scans. Richpanel, invisible at 0 of 15 in its own July 2026 scan, appeared in 4 of 14 question source sets in the gorgias.com scan that same morning. Same market, same day, same instrument — different question framing, different visibility. The gorgias.com run asked Shopify-helpdesk questions, where the ecommerce listicles and competitor comparison pages that mention Richpanel rank; the richpanel.com run asked AI-first support questions, where they did not.

Where the answers actually come from

Both scans logged every cited source with its domain and category, and the pattern repeats across the two files.

Reddit was the single most-cited domain in each run: 13 of 70 citations in the gorgias.com scan and 10 of 75 in the richpanel.com scan. Comparison listicles and community threads together accounted for 29 of 70 cited sources (41%) in the gorgias.com run and 33 of 75 (44%) in the richpanel.com run. YouTube shows up in both, with 3 and 4 citations respectively.

Much of the rest of the citation graph is vendors writing about each other's category. In the gorgias.com scan, kustomer.com tied YouTube as the second-most-cited domain at 3 citations, while gorgias.com itself was cited only 2 times out of 70 — in its own market's answers. In the richpanel.com scan, fin.ai — the domain of Intercom's Fin agent — tied YouTube for second place at 4 citations, alongside one or two citations each for yuma.ai, vellum.ai, crisp.chat, and a spread of other vendor-owned blogs. When a vendor's comparison content ranks, that vendor gets to write the shortlist. This is the mechanism described in how AI assistants reshape category discovery: the assistant does not survey the market; it reads whoever ranked.

Same market, two different maps

Overlaying the two runs — with the standing caveat that two runs is all this is:

Zendesk is the constant. 7 of 14 source sets in the gorgias.com scan and 7 of 15 in the richpanel.com scan: the only company at roughly half in both files.
Gorgias travels well. 6 of 14 in its own scan, 7 of 15 in richpanel.com's scan — where it tied Zendesk and Intercom for the highest presence of the run.
The supporting cast swaps out. The gorgias.com run's mid-table is ecommerce-helpdesk names: Gladly and Front at 3 of 14, Re:amaze and Shopify Inbox at 2. The richpanel.com run's mid-table is AI-native names: Ada and Sierra at 4 of 15, Decagon at 3, Forethought at 2. Change the question framing and the market changes cast around a stable core.
Visibility is fragmented. 43 distinct company and product names in the gorgias.com run's table and 44 in the richpanel.com run's, with 27 and 31 single-appearance names respectively. In these two files, most of this market is visible to a buyer only through a single ranked page.

That instability is exactly why AI share of voice has to be measured against a specific question set rather than assumed to be one number a company owns. Richpanel's July 2026 figure is simultaneously 0 of 15 and 4 of 14, depending on whose questions you ask.

Methodology: what a scan measures

A Magrios scan measures presence in top-ranking public web sources for real buyer questions, at a point in time. Concretely, for each run reported here: fifteen generated buyer questions across eight intent types; the top five ranking public sources collected per question via the web-sources engine; three sources per question read in full; companies extracted with evidence spans at a minimum confidence of 0.8; share of voice computed as the share of question source sets in which a company appears.

What a scan is not: it is not market share, revenue, install base, or product quality, and presence is not causation — a company can rank in a question's sources for reasons unrelated to anything it did recently. Each run here is a single repetition; ranked results shift day to day, so single-run numbers carry sampling noise. Two run-specific notes from the stored metadata: the gorgias.com run generated 15 questions and analyzed 14, and the richpanel.com run logged one crawl failure (a failed fetch of its target's homepage) during company profiling, which did not affect the source measurement. Above all: this is a two-scan view of one market on one day — July 9, 2026 — not a census and not a trend.

Evidence stored

Evidence stored July 9, 2026: the gorgias.com run (06:23–06:31 UTC; 14 buyer questions analyzed, 70 cited sources) and the richpanel.com run (08:00–08:11 UTC; 15 buyer questions, 75 cited sources), each retained with per-question ranked-source lists, extraction evidence spans, citation categories, and full share-of-voice tables in the Magrios evidence store. Every number in this piece resolves to one of those two files.