Which RMM and service desk software do AI assistants name for MSPs in the UK? A GEO study
We asked four AI engines the questions a small managed service provider (MSP) would ask about remote monitoring and management (RMM) and about service desk and billing software, counted the brands and sources in 120 answers and probed 22 company websites with the names AI crawlers use.
AI, TELL ME! MSP Software Visibility Study, United Kingdom. By the AI, TELL ME! team. Published October 2026.
Short answer. Across 120 AI answers to questions from small MSPs in the UK, the monitoring platform recorded 17 software brands, and five of them (Syncro, ConnectWise, Datto, Atera and NinjaOne) account for 64.1% of all mentions. Which brands appear depends on the segment: NinjaOne is named in 60 of 60 remote monitoring answers and in 12 of 60 service desk and billing answers. Company websites supply 77.1% of the cited items (from 68.5% to 96.0% depending on the engine). The link between a brand being named and its own site being cited is weak (rank correlation 0.39). No website property measured shows a detectable difference between the brands the engines name and the five that appear in no answer.

A small MSP with about 15 clients asks an AI assistant which RMM tools it should compare, and the answer names a handful of software brands and links to a few websites. This study counts which brands and which websites those answers use, and then checks whether the websites behind the brands explain who gets named.
Both halves of this study ran on our own tools
AI, TELL ME! is a full-cycle AEO (answer engine optimisation) and GEO (generative engine optimisation) agency, and we operate our own monitoring and website assessment tools: the monitoring platform that records what each AI engine says and cites, and the AI Access Report that records what a probe using the engines' crawler names can reach and read on a website.
The monitoring ran in English with a UK geo target:
Two segments with 60 answers each: RMM software for MSPs, and service desk and billing software for MSPs, a category that vendors call PSA (professional services automation).
Four engines with 30 answers each: ChatGPT, Gemini, Google AI Mode and Google AI Overview.
Questions in the words of a small MSP, such as "We run a small MSP with about 15 clients and need a better way to monitor devices, handle patching, and connect remotely when users have issues. Which RMM tools should we compare?" in the RMM segment.
The questions ask which software fits an MSP's setup, so the ranking shows which brands the engines name for a shortlist. Twenty-two company websites were probed as well, 17 belonging to brands that appear in the answers and 5 to brands that never do.
Definitions: how to read the numbers in this study
A mention is one count of a brand in an answer, and one answer can carry several mentions of the same brand.
Answers naming a brand counts the answers in which the monitoring platform's brand extraction recorded the brand at least once.
A cited item is one cited domain in one answer, with repeats inside an answer counted, and answers citing a domain counts the distinct answers that cite it.
A crawler identity is the name an AI crawler announces when it requests a page. The probe used 14, 8 that fetch pages to cite and 6 that fetch pages for training.
Content extraction score, passage quality and structure score are the report's scores out of 100 for how much of a page a retrieval pipeline recovers, how well a page survives being cut into passages and how well its markup tells a machine what the page contains.
A page-based score is one that needs a page read: text without JavaScript, content extraction, passage quality and structure. Crawler access and speed are scored without one.
Five brands hold almost two thirds of all mentions in the UK answers

Figure 1. Share of voice across all four engines, 1,259 mentions. Brands below the line were named in no answer.
Syncro, ConnectWise, Datto, Atera and NinjaOne hold 807 of the 1,259 mentions (64.1%), more than twice the 29.4% that five of 17 brands would hold if the mentions were shared evenly, and only 39 mentions separate the first from the fifth (Syncro 186, NinjaOne 147). In practice, a small MSP asking an AI assistant meets the same few names most of the time. The next four brands hold 30.7% and the other eight share 5.2%. Seven brands appear in nine answers or fewer, and one appearance in 120 answers is a single event that places no brand in the ranking (Figure 2 shows each brand's answers by segment).
Table 1. The ten most-mentioned software brands, all four engines.
Brand | Mentions | Share of mentions | Answers naming the brand (of 120) |
|---|---|---|---|
Syncro | 186 | 14.8% | 82 |
ConnectWise | 168 | 13.3% | 83 |
Datto | 156 | 12.4% | 76 |
Atera | 150 | 11.9% | 80 |
NinjaOne | 147 | 11.7% | 72 |
HaloPSA | 116 | 9.2% | 64 |
SuperOps | 116 | 9.2% | 66 |
Kaseya | 83 | 6.6% | 60 |
N-able | 72 | 5.7% | 32 |
Pulseway | 23 | 1.8% | 14 |
The leader depends on how you count
Syncro leads by mentions, with 186 (14.8%). ConnectWise leads by answers naming the brand, with 83 of 120 answers against 82 for Syncro. ConnectWise leads on ChatGPT (eight mentions ahead of Datto) and on Gemini (one mention ahead of Atera), and Syncro leads on Google AI Mode and Google AI Overview. The same five brands are the five largest by mentions and by answers naming, so we read them as a group and name the measure with every claim about a leader.
The segment decides who appears

Figure 2. Answers naming each brand, out of 60 answers per segment: remote monitoring and management on the left, service desk and billing on the right.
NinjaOne is named in 60 of 60 RMM answers and in 12 of 60 service desk and billing answers, and HaloPSA shows the reverse pattern (6 and 58; Figure 2). N-able, Atera and Pulseway lean to RMM, and Syncro, ConnectWise and SuperOps lean to service desk and billing. Eight of the ten brands named in the most answers differ between the segments past a threshold corrected for ten tests. Datto (37 and 39) shows no detectable difference and Kaseya (23 and 37) falls short of the corrected threshold (p = 0.011). A single overall visibility figure would hide a split as wide as 60 and 12. Service desk and billing counts cover the brands the engines named in the monitoring platform's extraction, and all of these are presence rates inside a segment that carry no ranking.
The four engines rank the brands alike and differ in breadth

Figure 3. Share of each engine's brand mentions per brand, in overall order. Shade runs from white (0%) to black (20% or more) and a zero is printed as 0.0.
Rankings correlate between 0.75 and 0.93
The four engines put the brands in a similar order: the rank correlations between their mention counts run from 0.75 (ChatGPT with Google AI Mode) to 0.93 (ChatGPT with Gemini) across the 17 named brands, all six pairs at p < 0.001. Ten of the 17 brands are named on all four engines, and the five brands named on one engine only all sit in the tail of the ranking (MSP360, PSOhub and Desk365 on Google AI Mode, DeskDay on Google AI Overview and GoTo (LogMeIn) on ChatGPT).
Table 2. Engine profile: mentions, brands named, the share held by each engine's three most-mentioned brands, distinct domains cited and cited items.
Engine | Mentions | Brands named | Three most-mentioned brands combined | Distinct domains cited | Cited items |
|---|---|---|---|---|---|
ChatGPT | 452 | 11 | 43.6% | 41 | 274 |
Gemini | 270 | 11 | 41.9% | 33 | 251 |
Google AI Mode | 277 | 14 | 40.4% | 38 | 89 |
Google AI Overview | 260 | 13 | 41.9% | 100 | 453 |
All engines | 1,259 | 17 | 40.5% | 152 | 1,067 |
ChatGPT names more brands per answer
ChatGPT gives the longest shortlists: its answers name 6.6 brands each and make 15.1 mentions each, against 5.1 brands and 9.0 mentions in the answers of the other three engines (Mann-Whitney p = 0.00001 for brands and p = 0.00003 for mentions; 30 answers against 90). Answer length alone does not explain it, because Gemini's median answer is longer than ChatGPT's (5,089 characters against 4,269).
Some brands depend on the engine
Four of the ten brands named in the most answers differ across the engines past the 0.005 threshold. ConnectWise is named in 28 of 30 ChatGPT answers and in 14 of 30 on Google AI Mode, and Datto in 26 and 14. N-able is named in 17 of 30 ChatGPT answers and in 5 on each of the other three engines, and Pulseway never appears on ChatGPT or Gemini and appears in 8 and 6 answers on Google AI Mode and Google AI Overview. Syncro, NinjaOne, SuperOps and HaloPSA show no detectable difference across the engines (p between 0.34 and 0.72), and Atera (p = 0.038) and Kaseya (p = 0.020) sit between the plain 0.05 level and the corrected threshold. The data do not say why a brand is absent from an engine.
Company websites supply 77.1% of the cited items, and the share and the mix depend on the engine
Most of what the engines cite is a company website: 823 of the 1,067 cited items (77.1%), from 68.5% on Google AI Mode to 96.0% on Gemini. The cited items split by type of site as follows (Figure 4 shows each engine):
The own sites of the 17 named brands supply 36.5% (26.0% without syncrosecure.com), other software vendors 32.1% and service providers and consultancies 8.6%.
Review, comparison and directory sites supply 11.6%, community and social sites 8.2% and media 2.4%.
One search-engine page (0.1%) and three domains that could not be typed (0.6%) make up the rest of the 1,067 items.

Figure 4. Share of cited items by source type, all engines and each engine, 1,067 items on 152 domains. Company websites are the first three rows added together.
One domain supplies a seventh of all citations

Figure 5. The 15 most-cited domains (A) and the number of domains by how many answers cite them (B). Two service provider sites are labelled by type.
One site stands far above the rest: syncrosecure.com supplies 151 of the 1,067 items (14.2%) and is cited in 43 of the 120 answers (35.8%), and 68 of its items sit in 13 Google AI Overview answers. Own sites as a group are cited in 93 of the 120 answers, and once syncrosecure.com is set aside, the other own sites supply 238 of the remaining 916 items (26.0%) and the next domain, ninjaone.com, holds 55 of the 916 (6.0%).
Beyond that, many domains appear once: 85 of the 152 domains (55.9%) are cited in one answer only and 11 are cited in ten or more, while the ten most-cited domains hold 459 items (43.0%).
The source mix depends on the engine
Community and social sites such as Reddit and YouTube appear only on the two Google engines, at 18.0% of Google AI Mode's items (16 of 89) and 15.7% of Google AI Overview's (71 of 453), and on neither ChatGPT nor Gemini. ChatGPT leans on review and directory sites (24.8% of its items, against 1.6% of Gemini's), and Gemini leans on other software vendors (52.6%, against 21.3% to 28.9% on the other three engines). Own sites are between 31.1% (Gemini) and 45.3% (ChatGPT) of items on every engine. Google AI Overview accounts for 453 of the 1,067 items (42.5%) from 100 of the 152 domains, so whole-run source shares lean towards that engine. In plain terms, the sources behind an answer depend on the engine, and citations seen on one engine are a poor guide to another's.
Most company websites served AI crawler requests, and structure was the lowest page-based score
Ten of the 22 company websites served all 14 crawler identities with no robots.txt rule against them, and 17 of the 22 served all 8 identities that fetch pages to cite. The sites appear as Site A to Site V (A to Q for brands in the answers, R to V for the five that never appear), and no letter is tied to a brand, a share or a rank.

Figure 6. Readiness scores for Sites A to V, one AI Access Report each. Shade runs from white (0) to black (100) and not-measured cells are hatched.
Crawler replies are reported as counts
Six sites refused one training crawler by name, and the report says that refusing a training crawler does not affect whether a site can be cited (Table 3).
Table 3. Crawler replies by site, as counts (22 sites, 14 crawler identities each).
Reply pattern | Sites |
|---|---|
Served all 14 identities, no robots.txt rule against them | 10 |
Served all 14 identities, robots.txt disallows every crawler | 1 |
Served 13 identities and refused 1 training crawler by name | 6 |
Served some identities and refused the rest (allow list by name) | 2 |
Refused all 14 names and a browser-like request, served the probe's own name | 1 |
Refused every request or returned a 404 | 2 |
Four sites returned no page to the probe, and 31 of the 120 answers (50 cited items) cite at least one of their domains. A probe describes what a test client received from one address, and only server logs show what a live crawler receives.
Access and text sit near the ceiling, and structure is lowest
Crawler access and text without JavaScript have medians of 100, passage quality 96, content extraction 92.5 and structure 83 (range 19 to 97), the lowest of the four page-based scores (Table 4). The one access score below 98, a 20, belongs to the site whose robots.txt disallows every crawler while its server served all 14 names. No site reaches 80 on lab speed (median 51), which is lower than any of the other five medians.
Table 4. Score ranges across the sites measured. A site with no score for a measure is left out of that row, and "not measured" is never read as zero. Speed is Google's PageSpeed Insights lab score on mobile, averaged over four or five pages at 18 sites and taken on one page at the other four.
Score | Sites measured | Range | Median |
|---|---|---|---|
Crawler access | 17 | 20 to 100 | 100 |
Text without JavaScript | 18 | 73 to 100 | 100 |
Content extraction | 18 | 64 to 100 | 92.5 |
Passage quality | 18 | 79 to 98 | 96 |
Structure | 18 | 19 to 97 | 83 |
Speed (lab) | 22 | 32 to 76 | 51 |
Page signals differ widely between sites
Across the 18 sites read on 19 to 21 pages each:
JSON-LD sits on every sampled page at 11 sites, on some pages at 4 and on none at 3.
A machine-readable date sits on every sampled page at 5 sites and on none at 4, and the two signals do not travel together: one site has JSON-LD on all 21 sampled pages and a date on 9.
The extractor keeps between 19% and 79% of the visible text (median 52%).
An llms.txt file is verified present at 9 of the 22 sites, verified absent at 9 and unreadable at 4.
Five brands appear in no answer, and all five sites were readable
Barracuda RMM (a security-focused remote monitoring and management platform from Barracuda Networks), Auvik (network management software), Tactical RMM (a source-available RMM), Freshservice (the IT service management product of Freshworks, with an MSP mode) and SysAid (IT service management software) are named in none of the 120 answers, answer text included. All five sites were read on 19 to 21 pages: three served all 14 crawler identities and two refused one training crawler by name. Two of the five were measured on the host the probe reached (a parent company's site and a documentation site), so their page figures describe that host. None of the 12 measures (the 14 site measures without the two finding counts) differs from the named-brand sites past a corrected threshold of 0.004 (smallest p = 0.024). One of the five scores 64 on content extraction, below all 13 named-brand sites read on pages (lowest 79).
No measured site property is detectably linked to being named
None of the 28 Spearman rank correlations between the 14 site measures and the two outcomes, the mentions a brand receives and the answers that name it, passes the corrected threshold of 0.0018 on the named-brand sites (12 to 17 sites per test).
Two reach p < 0.05, and both involve the share of visible text the extractor keeps. It correlates negatively with mentions (rho −0.70, p = 0.0075, n = 13) and with answers naming the brand (rho −0.67, p = 0.012, n = 13), and it falls to rho −0.22 (p = 0.39, n = 18) once the five never-named sites join at zero. At 0.05, 28 tests give 1.4 passes by chance.
The four sites that returned no page have no page-based score or page signal, so the tests on those measures cover 13 of the 17 named-brand sites (12 for crawler access and 17 for speed). With 13 sites no coefficient supports a claim. In plain terms, the site scores in this study did not predict how often a brand is named, and 13 sites are too few to say more.
Being named in an answer and having your own site cited are weakly related

Figure 7. Answers naming each brand (left) and answers citing a domain whose name identifies the brand (right), out of 120.
Among the 17 named brands, the rank correlation between mentions and own-site cited items is 0.39 (p = 0.12, n = 17), which is weak and statistically unclear, and 0.27 without the leading brand (p = 0.31, n = 16). A brand's own site is cited in the same answer in 121 of the 657 brand-answer pairs (18.4%), so being named in most answers says little about whether the brand's own site is cited.
Atera is named in 80 answers and has its site cited in 3, and HaloPSA (64 and 1) and Kaseya (60 and 3) show the same gap, while Pulseway has no own domain among the 152 cited domains. In the other direction, DeskDay is recorded by the brand extraction in one answer and has its site cited in 9, none of them that answer. Syncro (186 mentions, site cited in 43 answers) and NinjaOne (147 mentions, 19 answers) are named often and cited often. With the five never-named brands added at zero the coefficient rises to 0.69 (p = 0.0004, n = 22), and that rise comes from the five pairs at zero.
Four checks a software vendor can run before commissioning optimisation
In this sample, visibility in AI answers varied by segment and by engine, and the site scores did not explain it. The results point to four checks a vendor can run on its own category:
Check visibility for each segment separately: NinjaOne is named in 60 of 60 RMM answers and in 12 of 60 service desk and billing answers.
Compare the engines one by one, because N-able is named in 17 of 30 ChatGPT answers and in 5 of 30 on each of the other three engines.
Look at the pages the answers cite in your own category: 85 of the 152 cited domains appear in one answer only while 11 appear in ten or more.
Repeat the measurement on a fixed set of questions: this study is a single reading, and a change in share of voice only counts when the questions have stayed the same.
A website audit comes after those four checks, because this sample of 22 sites found no website property that differs detectably between the brands the engines name and those that no answer names.
After an audit, the work moves from measurement to a plan
For a single brand, AI, TELL ME! runs the work that follows a study of this kind:
AEO Monitoring shows where a brand appears in AI-generated answers and whether it appears at all, with the competitors and cited sources behind each answer.
The GEO Report assesses the brand's website, content and external presence and shows what helps the brand appear in AI recommendations and what holds it back.
The AEO & GEO Roadmap turns the findings into a prioritised 90-day plan for technical, content and external presence work, in order of impact.
Website and content work gives AI systems a clear, accurate picture of what the brand offers, and is delivered with the brand's own team, fully outsourced or as a mix of both.
PR for AI search builds the brand's presence in the sources AI systems use to describe it, with a target list of publications and expert voices, published coverage and tracking.
Monitoring then repeats on the same prompt set.
The data cannot say whether copying the leaders works
No measure in this study shows that copying a leader's website would win mentions, and none shows that it would fail. Some leading brands have their own sites cited in many answers (Syncro in 43 and NinjaOne in 19), and three findings keep the question open:
Own sites supply 36.5% of the cited items (26.0% without syncrosecure.com), so more than three fifths of the sourcing sits on sites other than the named brands' own.
Atera is one example of a visible brand with a rarely cited site: it is named in 80 answers and has its own site cited in 3.
No measured site property differs detectably between the brands the engines name and the five that no answer names, and all five of those sites were readable; with 5 sites against 13 the test is small.
These findings against other published research
Three published analyses bear on this study: each is tied to tracking or SEO software (a co-founder's platform, a partner's tool and a vendor's own product), none is reported as a UK-only sample and none measures the same thing under the same definitions.
Kumar (preprint of 18 June 2026, by the co-founder of the platform that supplied the data) reached nearly the same company-website share as this study: about 78% of 149,912 citations, in a dataset of 102,025 AI responses about 102 brands on five engines, against 77.1% here. The split inside that share differs: 75.2% of Kumar's citations come from sites owned by other businesses and 2.9% from brand-owned sites, against 36.5% brand-owned here (26.0% without syncrosecure.com). The brand sets differ, with 102 tracked brands there and one B2B category here.
Overthink Group, an agency that ran its tracking through a partner's tool, analysed 250 B2B software categories on four engines (14 July 2026) and found that the source mix depends on the engine, with YouTube at 7.4% of Google AI Overview citations and, across all engines, the G2 network at 8.0% and Reddit at 1.4%. This study agrees that the mix depends on the engine and finds YouTube at 3.5% of Google AI Overview items (16 of 453), the G2 network at 1.0% of all items (11 of 1,067) and Reddit at 4.3% (46 of 1,067), in a different market with different categories and engines.
Ahrefs, an SEO software vendor, tested 1,885 pages that added JSON-LD (11 May 2026) and found a 4.6% fall in citations in Google AI Overviews and rises of 2.4% in Google AI Mode and 2.2% in ChatGPT, the last two statistically indistinguishable from zero. That test compares pages before and after a change and finds no clear gain from JSON-LD, and this study compares sites at one point and finds no detectable link between a site's JSON-LD share and how often its brand is named (rho −0.05 with mentions, p = 0.86, n = 13).
Frequently asked questions
Which RMM software do AI assistants name most often for MSPs in the UK?
In the 60 RMM answers, NinjaOne is named in all 60, Atera in 49 and Datto in 37. Across both segments, Syncro, ConnectWise, Datto, Atera and NinjaOne hold 807 of the 1,259 mentions (64.1%).
Do AI assistants name different software for remote monitoring and for service desk and billing?
Yes: NinjaOne is named in 60 of 60 RMM answers and in 12 of 60 service desk and billing answers, and HaloPSA in 6 and 58. Eight of the ten brands named in the most answers differ between the segments.
Which sources do AI assistants cite for MSP software?
Company websites supply 77.1% of the 1,067 cited items, from 68.5% to 96.0% depending on the engine: the own sites of the named brands 36.5% (26.0% without syncrosecure.com), other software vendors 32.1% and service providers 8.6%. Review sites supply 11.6%, community sites 8.2% and media 2.4%.
Do all AI engines name the same MSP software brands?
The four engines rank the brands in a similar order (rank correlations from 0.75 to 0.93) and differ in who leads and how many brands they name: ConnectWise leads on two engines and Syncro on the other two, and ChatGPT names 6.6 brands per answer against 5.1 on the other three.
Does a better-built website make a software brand more visible in AI answers?
In this sample no link could be detected: none of the 28 tests of site properties against how often a brand is named passes the corrected threshold, and the five brands named in no answer all had readable sites.
Methodology and limitations
This study in numbers: one category (software for MSPs) in one market (the United Kingdom), two segments, four AI engines, 120 answers naming 17 brands, 1,259 mentions, 1,067 cited items on 152 domains and 22 company websites probed. Citation figures cover the 112 answers that carried sources and answer-level citation counts are out of 120. Five brands that no answer names were added at zero.
The run is a single reading, so figures are directional, and another question wording could change the ranking. Each segment is read on its own, because presence differs between them.
Share of voice divides a brand's mentions by all 1,259 mentions, or by one engine's mentions in the engine view. Answers naming a brand are counted out of 120 (60 within a segment, 30 within an engine), and a citation percentage divides by the 1,067 cited items unless the text names another base. The 17 brands are those the monitoring platform's brand extraction recorded, and the study does not claim a complete list of brands for either segment.
Segment and engine comparisons are chi-square tests on the ten brands named in the most answers, with a Bonferroni threshold of 0.005. Eight brands pass the segment test (p < 0.0001, except Atera at 0.0005), and Datto (0.70) and Kaseya (0.011) do not. ConnectWise (0.0003), N-able (0.0004), Pulseway (0.0009) and Datto (0.0037) pass the engine test, and Atera (0.038) and Kaseya (0.020) do not. Rank correlations are Spearman coefficients, and brands per answer are compared with a Mann-Whitney test.
The 14 site measures in the 28 correlation tests are the six scores in Table 4, five page-signal shares (JSON-LD, machine-readable date, one H1, unskipped heading levels, title and description), the extractor's kept-text share and two counts of findings in the report. No test survives the 0.0018 threshold.
Source types are our classification of each cited domain into eight groups by its owner's business (Figure 4), and three domains (6 items, 0.6%) could not be typed. A cited domain is a brand's own site when its name identifies the brand, and sites of product brands under another domain are not mapped.
Each of the 22 company sites received one AI Access Report: 14 crawler identities plus two robots.txt permission tokens, sent to the home page from one address and compared with a desktop browser request, plus up to 20 sitemap pages across the page templates. Eighteen sites were read on 19 to 21 pages and four returned no page. The probes ran after the monitoring run on global company domains, with no UK network, and scores are a relative scale for tracking change that is not calibrated against citation.
Five of the 22 reports describe a repeat run on a later day, the page samples for four of those five were drawn again and one other sample spans four languages, so each site figure describes the sample drawn.
Access findings carry one of three evidence grades: site-attributable (the evidence sits in a page's HTML or a static file, or a server answered some crawler names and refused others inside one run), withheld (every request was refused or returned a challenge page) and contradicted (the finding sits against other evidence in the data). Twenty sites are site-attributable and two are withheld, and three sites also carry the contradicted grade. Four sites returned no page, two of them among the 20.
Site letters were assigned by a seeded shuffle. The five brands that no answer names were chosen by us from vendors' own pages and press reports and fit the category to different degrees, since Auvik is network management and Freshservice and SysAid are IT service management products. Brands are spelt as each company writes them, and GoTo (LogMeIn) carries both names the company uses.
Running a GEO study for your own category
This study covers software for MSPs in the UK, and the same measurement runs for any category, in any market and language, on the same two halves: what the engines say and cite, and what a probe using their crawler names can reach and read.
We bring 17 years in marketing, PR and SEO to this work.
Book a 30-minute review of your own category at https://calendly.com/natali-khld/welcome.
Request the prompt set, the engine-by-engine breakdown or the AI Access Report for your own website at hey@aipleasetellme.com.
Written by the AI, TELL ME! team (aipleasetellme.com), October 2026.


