Gorgias Review 2026: Helpdesk Pricing and AI Costs

The current price table has five Helpdesk base rows. Starter is $10 monthly with 50 tickets and three seats. Basic is $60 monthly or $50/month billed annually with 300 tickets; Pro is $360 or $300 with 2,000; Advanced is $900 or $750 with 5,000. Enterprise is custom-priced. Starter has no verified annual equivalent. Current pricing also advertises AI Agent on every plan and an included 30-interaction Starter allowance, while the setup guide requires an active AI subscription. Confirm the exact offer rather than assuming either no included AI or unlimited automation.

The website specifies 500 seats for Basic, Pro, and Advanced, while the Shopify listing describes them as unlimited. Use the exact account entitlement rather than interpreting ticket-based pricing as unlimited capacity in every dimension. BigCommerce is gated from Basic; Magento and revenue attribution from Pro. SAML appears with custom plans, so identity requirements can change the appropriate starting configuration.

Price comparisons must keep their meters consistent. A Zendesk shortlist should compare the required agents, channels, and AI usage against Gorgias's ticket volume, not merely two headline amounts. A different AI-inclusive quote does not invalidate the $10 Starter base, and a small included allowance does not remove automation limits or overages. Request the base subscription, included interactions, active AI entitlements, and additional-usage charges explicitly before approving the purchase.

The billing guide defines the event: a human, Rule, or AI Agent sends a message. Several messages within that ticket do not each create another ticket charge. An automatic acknowledgment therefore deserves the same billing review as an agent reply; the rule's helpful intention does not change the fact that it sends an outgoing message.

Returning conversations also matter. A later reply can create a new ticket after more than ten days of email inactivity or more than three days for chat. These are distinct from the AI handover window. A seasonal retailer should model repeat contacts and delayed follow-ups, not assume that one customer's original order creates only one billable support ticket indefinitely.

Unanswered spam and newsletters are not billable, but replied-to messages can be. The Voice spam guide also counts answered spam calls and spam voicemails. Review channel-specific behavior and automation rules when following our SaaS cost-control guide. A spam label is not a billing exemption, and a reduction in manual replies is not necessarily a reduction in Helpdesk ticket consumption.

Under current billing, a conversation resolved entirely by AI can consume one Helpdesk ticket and one automated interaction. These are not mutually exclusive counters. If the AI sends the reply and completes the conversation without human involvement, removing that ticket from a Helpdesk forecast would understate usage. Budget the ticket pool and the automation pool separately before predicting savings from deflection.

Human involvement within 72 hours prevents the automated-interaction charge, while the Helpdesk ticket remains billable. Other self-service features, such as a Flow resolving without a Helpdesk reply, can instead incur automation only. The AI analytics definitions distinguish closed tickets from billable automated outcomes and describe the 72-hour reporting delay. An immediately closed ticket is not a finalized invoice metric.

There is an important legacy-pricing exception: eligible older accounts use either ticket or automation charging. A screenshot of another merchant's legacy bill does not establish a new buyer's terms. Confirm account eligibility and keep that evidence with the quote. This distinction explains why apparently similar descriptions of Gorgias billing can lead to different totals without either describing the same subscription.

Consider Basic on monthly billing with 400 billable tickets. Its $60 base includes 300, and an additional 100 costs $40: $100 before AI, add-ons, and tax. If 50 of those conversations were AI-only resolutions, the Helpdesk count remains 400, with 50 automated interactions additionally consumed. It does not fall to 350 simply because an employee did not send those replies.

The pricing FAQ quotes $0.90 annually or $1 monthly per automated interaction and $1.50 overage, but the Advanced comparison grid advertises $0.85. Those statements do not support one universal AI rate. Confirm the plan, commitment, allowance, and applicable overage together. Basic's billing guide explicitly charges $0.40 per additional ticket: 302 tickets cost $60.80 and 400 cost $100 before AI, add-ons, and tax. Confirm other tiers' applicable overage terms rather than treating every displayed per-100 rate as a rounded block purchase.

Build quiet, normal, and peak-month forecasts with the same billing definitions, then price the complete quote. Our software evaluation framework helps separate required features from optimistic usage assumptions. Do not annualize only the base price while ignoring seasonal tickets, paid channels, and automation. These examples are arithmetic based on published rules, not an observed merchant bill or a promised cost reduction.

The AI Agent setup guide requires a connected Shopify store, an AI subscription, and Lead/Admin/Owner permissions. It documents knowledge sources, skills, configurable actions, human handover, and email, chat, SMS, and social deployment. Those capabilities should be evaluated against the current store configuration; another platform's Helpdesk connector does not independently meet the Shopify requirement.

Knowledge and actions need named owners. A useful evaluation includes a clear policy question, missing order information, a request outside policy, and a case requiring escalation. Check which action is permitted and whether the handover supplies enough context for a person to continue. Our customer-service AI chatbot guide explains why a correct-looking answer is not sufficient evidence of safe automated service.

Do not optimize solely for the displayed automation percentage. Review whether the customer received a correct answer and whether a later human intervention changes the billing outcome. Teams also evaluating Chatwoot should compare deployment responsibility, AI meters, and support-channel needs independently. No AI model, refund workflow, live order action, or conversion result was benchmarked in this documentary review.

Gorgias does not have a verified ongoing free plan. Pricing and the Shopify listing advertise seven days, but the September-updated account-creation guide says 30 days. Report the discrepancy rather than combining the longer duration with whichever other benefits seem preferable. The signup offer and written confirmation should determine the trial available to the actual account.

That creation guide also limits outbound email to 20 per day, 150 per week, and 600 across the trial, and prohibits creating a Help Center. It therefore cannot support an unrestricted-all-features claim. Ask whether the proposed evaluation can exercise the necessary channel, knowledge source, permissions, AI subscription, and export route before scheduling a complete customer-support migration around temporary access.

If the evaluation's scope or budget is unsuitable, compare Freshdesk and Crisp using the same channel and workload requirements. A different entry price is not proof of a cheaper equivalent system. Keep trial findings separate from production promises, particularly where expected automation rates or staffing savings would otherwise determine the purchase.

The ticket export guide separates two routes. View CSVs contain metadata and links for up to one million tickets. The message-content analytics export covers 30 days and up to 100,000 tickets, includes public messages only, and excludes internal notes and specified social-media content. Download links expire after 14 days. A large row allowance does not remove these content exclusions.

Permissions differ by route. The export guide specifies Owner/Admin/Lead access, while the newer permissions matrix permits analytics CSV downloads across roles. Do not convert this into a blanket claim that every CSV is administrator-only. Have the intended export operator verify the relevant pathway, then inspect representative messages and missing notes before promising a complete migration archive.

The subscription agreement, section 8.2, provides requested customer-data availability for 45 days after termination. Pricing's reference to deletion within 90 days addresses a different window; it does not guarantee account access for 90 days. Export required records before cancellation, obtain clarification for excluded data, and secure any downloaded customer information according to the business's actual retention and access obligations.

Gorgias deserves consideration when ecommerce order context and the documented store integrations are central to support work. It is less compelling to choose it simply because AI is prominently advertised or the Starter headline is low. Our customer-support software directory offers alternatives for different workflow priorities; the right comparison is the complete operational configuration, not an isolated feature or entry price.

For current buyers, the decisive checks are Shopify eligibility for AI, the exact billable-ticket event, dual metering, the 72-hour human-involvement boundary, and realistic seasonal demand. Resolve the trial and rate-card discrepancies in the quote. Confirm the required permission roles and whether available exports meet your archive obligations. These are purchase conditions, not minor details to discover after an annual commitment.

Proceed with a scoped evaluation if those boundaries fit, with named owners for knowledge, actions, billing, and migration records. Skip an AI-led purchase if eligibility is unresolved or the business requires a complete archive that the documented export routes cannot provide. This conclusion is documentary and conditional: no inbox throughput, support speed, customer satisfaction, or AI resolution performance was measured.