The UK Accounting Software AI Visibility Index 2026

Why we built this

Buyers increasingly use ChatGPT, Gemini, and Google AI Overview alongside traditional search when researching software purchases. Yet almost nothing has been published about which brands those systems actually recommend, how consistently they agree with each other, or where they get their information from.

We picked one category, UK accounting and bookkeeping software, and tested it properly. Twenty real buyer questions. Three AI systems. Every response captured in full, scored on the same framework, and checked against what each engine actually cited as its source.

This is the first release in a wider UK AI Search Visibility Index.

Methodology

• Category: UK accounting and bookkeeping software, chosen as a pilot niche because it carries no regulatory sensitivity, has strong existing AI training coverage, and maps to clear, non-branded buyer intent.
• Brand universe: 20 UK accounting software providers, including Xero, QuickBooks Online, FreeAgent, Sage, Zoho Books, Pandle, QuickFile, Wave, KashFlow, and others that surfaced organically in responses.
• Prompts: 20 non-branded, commercially meaningful buyer questions, spanning general comparisons, persona-specific queries, compliance-driven queries, and direct comparisons.
• Engines tested: ChatGPT, Google Gemini, and Google AI Overview, each prompt run once per engine in a fresh incognito session with a new chat per prompt.
• What we recorded: whether each brand was mentioned, whether it was actively recommended, its position within the answer, and every source domain the engine cited.
• Tie handling: where an engine presented two brands as joint top picks, each brand is credited with half a win in every percentage calculated in this report, applied consistently throughout.
• Scope note: each prompt was captured once per engine rather than averaged across repeated runs. This is a controlled pilot methodology, not a claim of permanent market position.

01 Xero wins the AI recommendation race, and wins it by a wide margin

Xero was the top pick in 70 percent of ChatGPT prompts (14 of 20, no ties), 60 percent of Gemini prompts (11 outright wins plus two ties with QuickBooks, weighted), and 73 percent of Google AI Overview prompts (14 outright wins plus one tie with QuickBooks, weighted). No other brand came close on any single engine. The nearest competitor, FreeAgent, topped out at 25 percent (5 of 20) on its best-performing engines.

What is notable is not just that Xero wins most often, but where it wins. It dominates every prompt framed around growth, scale, multi-user collaboration, or "which software do accountants recommend to clients." Its losses cluster almost entirely around one specific persona: the simple, budget-conscious sole trader.

02 FreeAgent's wins are not random, they follow a persona rule

FreeAgent was the only brand to ever take the number one spot away from Xero on a consistent basis, and it did so in a highly predictable pattern across all three engines. Every single FreeAgent win came on a prompt containing one of these signals: "sole trader," "freelancer," "contractor," "invoicing," or "free."

The moment a prompt shifted toward growth, multi-user access, inventory, or e-commerce complexity, FreeAgent dropped out of the top three entirely, in some cases falling to last place among the five brands typically surfaced.

This is a genuinely useful finding for FreeAgent's own marketing: its AI visibility is not broad, it is precisely targeted to one buyer persona, and it earns that position consistently across independently trained AI systems that have no shared knowledge of each other's outputs.

03 QuickBooks is the most mentioned, least recommended brand in the dataset

This is the most counterintuitive finding in the report. QuickBooks Online appeared in every single one of the 60 responses across all three engines, a 100 percent mention rate matched by no other brand. Yet it was never once the outright, unqualified top recommendation from ChatGPT across 20 prompts, and only tied for top spot once in Google AI Overview. Gemini gave it the top spot once outright and twice in a tie with Xero.

In practical terms, AI systems treat QuickBooks as a name that must always be mentioned in any serious comparison, a kind of default inclusion, but rarely as the name they lead with. QuickBooks has near-total AI awareness without AI preference, and closing that specific gap looks like the more useful strategic target for the brand.

04 The same brand can be nearly invisible on one AI engine and dominant on another

Sage was mentioned in 19 of 20 ChatGPT responses and 18 of 20 Gemini responses, a near-universal presence, consistently as a strong runner-up on payroll-heavy, CIS, or established-business prompts. In Google AI Overview, it was mentioned in just 5 of 20 responses, and only appeared with any real substance when a prompt explicitly named it.

This is a structural visibility gap tied entirely to which AI surface the buyer happens to use, and it is exactly the kind of finding that a brand's own SEO and AI-visibility strategy needs to know about and actively address.

05 Nine established brands, zero AI visibility

Of the 20 well-known UK accounting brands we set out to track, 9, Crunch, Coconut, Countingup, Bokio, Tide, Nomisma, Capium, IRIS, and Digita, never appeared organically in a single one of the 60 real AI responses collected. That is 45 percent of the tracked brand universe receiving zero organic AI visibility across every prompt we tested.

These are not obscure companies; several have meaningful UK market presence and real customer bases. It suggests that for a large share of established software providers, current AI search behaves less like a long tail and more like a closed shortlist of four or five names, repeated with remarkable consistency across independently trained systems.

06 Google AI Overview showed the highest rate of citation mismatch in our test

ChatGPT relied heavily on GOV.UK for compliance-flavoured prompts and on the brands' own websites for commercial comparison prompts, with occasional independent third-party sources.

Gemini cited a noticeably wider range of independent UK sources, including PCMag UK, Airwallex, ReconcileIQ, Outbooks, and multiple named UK accountancy practices, generally with clear topical relevance.

Google AI Overview cited by far the widest and most varied source mix of the three, including real UK chartered accountancy firms, comparison blogs, a LinkedIn post, and two YouTube videos. It was also the only engine to show a repeated pattern of citing sources that did not actually match the query: roughly one in three prompts included at least one citation that was off-topic, geographically mismatched, or reused from an unrelated query. Nephos Group's "8 Best Accounting Software" listicle was cited as a source four separate times regardless of what was actually asked.

A handful of domains recurred across more than one engine: Startups.co.uk, Snyp.ai, outbooks.co.uk, and Airwallex each appeared in both Gemini and Google AI Overview, and 123Financials was the only domain cited by all three engines.

07 AI confidence does not always match the evidence it cites

On the prompt "what accounting software do UK accountants recommend most," ChatGPT surfaced a real 2024 survey, noted it was commissioned by QuickBooks and should be treated with caution, reported the survey actually put QuickBooks first at 36 percent against Xero's 32 percent, and then still concluded Xero was the safer recommendation on ecosystem grounds.

Google AI Overview, asked the identical question, stated flatly that Xero is "the absolute most recommended" software by UK accountants, with no hedge and no reference to any conflicting data.

Same question, same underlying category, two very different levels of epistemic caution. The confidence with which an AI system states a claim is not a reliable signal of how well-supported that claim actually is.

08 AI recommendation strength does not track review footprint

Sage has the largest Trustpilot review footprint of any brand in the dataset, roughly 20,000 UK reviews. Its average AI recommendation rate: effectively zero. QuickBooks sits second by review footprint, around 17,000 reviews, and also converts almost none of that into being an AI system's top pick.

Xero, by contrast, has a smaller footprint than either, around 11,200 reviews, yet converts that into a 68 percent average AI recommendation rate. FreeAgent shows the same pattern at smaller scale: roughly 2,900 reviews, but the second-highest AI recommendation rate at 23 percent.

Whatever is driving an AI system's decision to lead with a particular brand, it is clearly not simply which brand has the most reviewed footprint.

Objective signals checked against the AI data

Wikipedia and Wikidata. Xero, QuickBooks, Sage, Zoho, and Wave all maintain established Wikipedia articles. FreeAgent has a full article and a detailed Wikidata entry, including its 2018 acquisition by NatWest Group. Pandle and QuickFile returned no standalone article in our checks.

Google organic search. Spot-checking two representative prompts showed independent listicle and comparison sites dominating organic results, with Xero, QuickBooks, Sage and FreeAgent all appearing as named winners depending on the site. Two separate PR-newswire style articles surfaced, one crowning Sage and another QuickBooks, both traceable to paid press-release distribution rather than independent editorial judgement.

The Netsleek AI Visibility Score

A simple, published-formula score for the five brands with reliable, fully cross-checked data across all three engines.

Pandle, QuickFile, and Wave are excluded from the scored table; each appeared in the dataset but not with the same full, cross-checked mention data across all three engines.

The AI Preference Gap

We define the AI Preference Gap as Mention Rate minus Number-One Rate: a brand every engine already knows about but rarely leads with will show a wide gap.

QuickBooks sits clearly alone at the top of this table. Sage and FreeAgent land within two points of each other but for opposite underlying reasons: Sage is mentioned less often and never converts; FreeAgent is mentioned almost universally and converts narrowly within its persona. Xero has by far the narrowest gap, meaning it converts the highest share of its mentions directly into being the top pick.

Recommendation Conversion Rate

Expressed as a ratio instead of a difference: of the times a brand was mentioned at all, what share of those mentions converted into it actually being the top pick?

Xero converts roughly seven of every ten mentions into the outright top spot. QuickBooks, despite being mentioned in every response in the dataset, converts fewer than one in twenty.

Competitive displacement: who replaces Xero when it doesn't win

Across the 60 engine-prompt instances in this dataset, Xero was the outright or shared winner in roughly 40.5, leaving about 19.5 instances where it was not. FreeAgent picked up 14 of those 19.5, roughly 72 percent. Every other brand combined accounted for the remaining 28 percent.

FreeAgent is Xero's only meaningful AI-search competitor in this category. When Xero loses the top spot, it is overwhelmingly FreeAgent stepping into it, not Sage, not QuickBooks, and not any smaller specialist.

Engine Dependency Score

Cross-Engine Consistency (used above) is based on Number-One Rate spread, which has a blind spot: a brand that never wins anywhere, like Sage, scores as perfectly "consistent." Engine Dependency instead measures the spread in Mention Rate, how dependent a brand's basic visibility is on one specific engine.

Sage has, by a wide margin, the highest engine dependency of any major brand tested: a near-default mention on ChatGPT and Gemini, almost absent from Google AI Overview.

Prompt coverage: which brand owns which type of buyer question

Grouping the 20 prompts into the four buyer-intent clusters used to design the prompt set, and calculating each brand's weighted share of engine-slots in that cluster, shows where each brand's strength actually sits.

Xero's weakest relative showing is still the general/comparative cluster, where FreeAgent claims a full third of visibility. Xero's strongest showing is the buyer-decision cluster, where it captured every single engine-slot with no exceptions.

Most frequently cited sources

123Financials is the only domain confirmed cited by all three AI engines. Nephos Group's four citations all traced to the same generic listicle regardless of the prompt asked.

Owned versus earned. Roughly a quarter of citations came from the brands' own websites, a further tenth from GOV.UK, and well over half from independent third-party sources. Most of what AI systems cite when recommending UK accounting software is written by someone other than the software company itself.

The Netsleek AI Visibility Framework

Ten of eleven dimensions measured with data we can defend in this release. Corroboration is deliberately left open: we did not tag every citation to the specific brand it was supporting within each response, and publishing invented per-brand corroboration figures now would repeat a precision problem this report has already had to correct once. That is the next concrete piece of work, not a gap to paper over.

What the findings suggest about how AI systems select a brand

This section is interpretation, not a measured result, and should be read that way.

One pattern runs through nearly every finding in this report: the three engines converge on the same four-way persona split almost every time, independently of each other. Xero for growth and scale. FreeAgent for sole traders and simplicity. Sage for payroll and complexity. QuickBooks for automation and ease of use. This is the framing ChatGPT, Gemini, and Google AI Overview each produced on their own, with no knowledge of each other's answers, and it shows up almost word for word, response after response.

One plausible explanation, and the one we find most consistent with the rest of the data, is that all three systems are drawing on a shared, pre-existing genre of "best accounting software for X" content that already exists across the web in exactly this structure, rather than each system independently reasoning from technical specifications. We want to be explicit that this is our interpretation of the pattern, not something the experiment directly proves.

If this interpretation holds up under further testing, the practical implication is worth stating even provisionally: the lever available to a brand may not be "explain your product better to an AI." It may be "become the clear, repeated answer to a specific buyer persona across independent third-party content," a content and positioning problem rather than a technical one.

What the data can't yet tell us

A genuinely deeper version of this question would need each brand's domain authority, referring-domain count, years established, brand search volume, and independently counted editorial mention volume, checked for statistical correlation against AI recommendation strength. We have deliberately not done that analysis in this release. That is strong candidate work for a dedicated follow-up study, not something to bolt onto this pilot.

Engine-specific leaderboards

Brand snapshots

Limitations

This is a controlled pilot, not a definitive market-share study. The figures reflect a single, dated capture per engine per prompt, not an average across repeated runs. The brand universe was fixed at 20 well-known providers going in, though every engine also surfaced additional brands organically. The objective-signal layer was collected for the brands that actually appeared as consistent AI recommendations, and the Google organic check covered two representative prompts rather than all 20. Full third-party press coverage volume, domain authority, and backlink data were not assessed.

Appendix: prompt-by-prompt results

The outright number-one recommendation from each engine, for all 20 prompts.

Xero won or tied for the top spot on at least one engine in 15 of the 20 prompts, and was the unanimous, outright top pick across all three engines simultaneously in exactly 10 of the 20 prompts. FreeAgent was the unanimous top pick across all three engines in 3 of the 20 prompts (sole traders, invoicing freelancers, and contractors), and came close on two more where two of the three engines agreed but the third named a different brand.