Travel eSIM Customer Service & AI: The 10 Largest Providers Compared

Jetpac leads with a 4.8 rating, followed by Saily at 4.7 and Holafly and Maya Mobile at 4.6. Telecom arms cluster at the bottom, with Matrix Cellular at 3.0 and Airalo at 3.9 after a trough of roughly 2.6 in mid-2025.

Ratings have been broadly stable-to-rising since 2024, with one notable exception: Airalo dipped sharply during mid-2025 before recovering. Review volume has grown for every consumer brand (Holafly ~55k→81k, Saily ~4k→24k), consistent with heavy invitation-prompted inflows, while scores held or improved — suggesting that the volume growth is not diluting quality.

The providers whose reviewers raise AI most often are also the ones with the largest 1–2★ tails — Airalo sits at ~9–10% AI mentions and a 25% low-rating share, Jetpac at ~3% and 5%. On ten providers, eight of them modelled, that is an association worth noticing rather than a demonstrated cause, and we have not tested any of these support flows end to end.

Each provider name links to the Trustpilot profile its score was read from, so you can check the current number against ours. Scores were read on 18 June 2026 and are current only as of that date — several have moved since, which is exactly why they are linked. AI-mention rates are hand-counted for two of the ten providers (Airalo, Holafly) and modelled for the other eight from support-channel signals; the Basis column says which is which, row by row. * Telecom arms have no standalone eSIM consumer profile: Bouygues and Matrix link to their main company profiles, and Vodafone to its Travel eSIM profile, which carries far fewer reviews than the flagship Vodafone profiles. Treat all three as proxies rather than like-for-like with the eSIM natives. Nothing in this table is an audit of any provider's support system.

All ten providers now run some form of automation. Our estimates of how much each resolves without a human span roughly 35% to 75%, but only one provider's stack is confirmed (Nomad, on Intercom Fin) and the rest are inferred from support-channel signals and hands-on testing — so read the spread as directional. The variable we think matters is whether the AI acts in backend systems or only retrieves text and hands off.

Strong in-bot resolvers — Holafly, Saily, and Maya — sit well above the deflect-then-human group. Nomad is the only provider with a confirmed vendor (Intercom's Fin); the rest are informed estimates based on support-channel signals, app-store reviews, and hands-on testing of each provider's live support flow. All resolution rates are benchmarked against 2026 industry data: enterprise tier-1 deflection median ≈ 41%, agentic in-bot resolution 70–85%.

The heavy-AI group (Airalo, Holafly, Saily, Nomad) shows the widest spread — from 50% to 75% — precisely because presence alone does not determine outcome. Airalo and Holafly both run heavy automation, and our estimates put them at opposite ends of the spread. We have not measured either system directly, so treat the gap as a hypothesis about in-session resolution versus hand-off, not a finding about either product.

Resolution = estimated share of contacts resolved through the AI-assisted layer without unaided human effort, bucketed into High (70–85%), Med (55–70%), and Low (40–55%) bands. Holafly anchored in the High band (industry baseline); all others are informed estimates.

Airalo's rating crashed to roughly 2.6 in mid-2025 — the most severe decline across any consumer eSIM provider — because its AI bot in WhatsApp contains tickets without resolving them, then hands a frustrated customer to a human who restarts the troubleshooting from zero. That single failure pattern accounts for a disproportionate share of its 1–2★ reviews and the elevated ~9–10% AI-complaint rate.

The same dynamic, at lower intensity, appears at Nomad (~11-minute waits to reach a human after the bot stalls) and Vodafone (TOBi deflects to call centre). In each case, the AI is measured by ticket containment, not ticket resolution — and customers can tell the difference.

Airalo's partial recovery to 3.9 by mid-2026 reflects improved routing rather than a resolved architecture problem: reviews cite faster escalation but the same bot pattern. The 25% low-rating share remains the highest in the top 10, more than double Holafly's 8% and five times Jetpac's 5%.

The implicit lesson from the benchmark is structural: containment-first AI produces short-term ticket deflection metrics that look good internally while visibly degrading CSAT and review scores externally. Resolution-first AI — where the bot takes real action in backend systems — produces the opposite effect.

The gap is architectural: deflection bots retrieve answers; resolution bots take actions. Closing that gap requires multi-agent orchestration connected to backend systems — re-provisioning a dead eSIM, issuing a refund, switching a plan — not a smarter FAQ engine. aissist.io is an AI operational layer (AgentMesh for orchestration, Pulse for insight, Evolve for optimization) designed to sit over a provider's existing helpdesk rather than replace it, eliminating the deflect-then-human loop that drives low ratings.

The benchmark exposes one consistent failure mode: a bot that contains the ticket but cannot act, then hands a frustrated customer to a human who restarts from zero. That pattern depresses CSAT, concentrates AI complaints in the 1–2★ tail, and produces the exact gap between Airalo (50% resolution, 3.9 rating) and Holafly (75% resolution, 4.6 rating).