Gong AI credits are the usage meter Gong introduced on June 2, 2026 for its at-scale AI features. Each paid core seat includes roughly 2,000 credits per year, pooled across the company and reset each contract year. Credits are consumed by automated AI processing — AI Trackers, MCP server tools, and API-based AI workflows — while interactive everyday use stays unmetered. When the pool runs out, credit-consuming jobs pause until the next contract year or until more credits are purchased.
That is the whole model in one paragraph, and to Gong's credit it is more transparent than surprise overage billing. But if your team built its workflows in the years when Gong's AI was effectively unlimited, the meter changes the economics of a product you already budgeted for — and most teams will discover exactly how at renewal time. This guide is the plain-English version, sourced from Gong's own documentation and independent analyses.
Gong AI credits are a pooled annual allowance (~2,000 per paid seat) consumed by Gong's automated AI processing; when they run out, those jobs pause until renewal or a top-up purchase.
What changed on June 2, 2026
Before mid-2026, Gong's AI features were effectively included in the seat and platform price. The credit model introduced a usage dimension: the company-wide pool is funded by seats (roughly 2,000 credits per paid core seat per year, per Gong's help center), and specific categories of AI work draw it down. Gong provides admin controls for allocating credits and guidance on optimizing AI Trackers to consume less — which tells you something about which feature drives the burn.
What consumes credits — and what doesn't
| Consumes credits | Does not consume credits |
|---|---|
| AI Trackers (Smart Trackers) run at scale across calls | Asking the Gong assistant questions by hand |
| MCP server tools | Interactive coaching workflows |
| API-based AI workflows, including the briefs API | Forecasting |
| Web search in briefs | Running a brief manually |
The pattern: interactive use is free; automated, at-scale AI processing is metered, and consumption scales with the volume of data processed. A tracker sweeping every call in a 100-rep org processes a lot of data. (Sources: Gong docs, GoNimbly's analysis, Discera's breakdown.)
What happens when credits run out
Per Gong's documentation on managing credit usage, credit-consuming jobs stop when the pool is exhausted, and resume when it resets at the new contract year or when additional credits are purchased. The good news: no surprise overage invoice. The operational news: whatever automated analysis your team relies on — tracker alerts, API-fed dashboards, automated briefs — quietly stops producing output, potentially months before renewal.
Why tracker-heavy teams burn fastest
The teams most exposed aren't doing anything exotic. They are running the same setups they built when the AI was free: broad Smart Tracker libraries sweeping every recorded call, API workflows feeding other systems, automated briefs on every account. Independent analyses of the model (GoNimbly, Discera) reach the same conclusion Gong's own optimization guidance implies: the allowance covers normal interactive use comfortably, and it is the automated, always-on configurations that outrun it.
What this means for your renewal
Credits land on top of the pricing structure buyers already negotiate: per-seat licenses, a platform fee, and auto-renewal uplifts that procurement analyses commonly report in the 5–15% range (see Tropic, Sybill's pricing breakdown). Four questions worth asking before you sign:
1. What did we actually consume? Ask for a credit usage report by feature. You cannot negotiate what you cannot see.
2. Which automated workflows earn their burn? A tracker that nobody's dashboard depends on is pure cost.
3. Is the credit allowance in the contract? Allowance size, top-up pricing, and rollover terms are commercial terms, not fixed facts.
4. What are we paying the AI to tell us? This is the bigger question, and it deserves its own section.
The question underneath the credits question
Metering forces a useful discipline: you now have to decide which AI analysis is worth paying for by the unit. Keyword trackers tell you whether words were said. Briefs summarize what happened. Those have value — but neither tells you the thing revenue leaders actually want to know: is my team selling well, and which specific behaviors are creating or destroying revenue?
That is a different measurement layer. In the stack we describe across this site: your CRM tells you what was logged. Conversation intelligence — the category Gong defined — tells you what was said. Revenue intelligence tells you what is happening in the pipeline. Revenue Behavioral Intelligence™ tells you which sales behaviors are creating or destroying revenue.
To be clear about what we are and aren't: Elanor is not a Gong replacement — it doesn't do forecasting, deal boards, or engagement. It scores every recorded sales call against rubrics built for the specific meeting type and produces the RBI Score: one decomposable number for the quality of your team's selling behavior, with the evidence timestamped on the call. Some teams run it alongside Gong; teams that mainly wanted evaluation and coaching run it with the recording stack they already have. If the credit meter has you auditing what your AI spend actually measures, that audit is worth extending one layer deeper — we'll run it on your own recordings.
Frequently asked questions
Sources: Gong Help Center (About Gong credits, Managing credit usage, Optimize AI Trackers); GoNimbly; Discera; Sybill; Tropic. Details reflect public documentation as of August 25, 2026; Gong's terms may change. Gong is a trademark of Gong.io Inc.; Elanor Labs is not affiliated with Gong.