Invoca’s July 13 benchmark report shows that calls originating from ChatGPT are considerably more likely to be qualified as sales leads than those from any other channel we track. When the call is answered, the conversion rate aligns closely with the overall average.
The data reveal a 49% lead rate for ChatGPT‑referred calls—about ten points higher than the 39% average across the seven channels Invoca monitors and six points above Google Business Profiles, which sit at 43%. While lead generation is strong, the subsequent conversion rate of roughly 40% is essentially on par with the overall 42% average that Invoca deems typical.
These figures are drawn from Invoca’s own customer base, which includes more than 70 million calls and 600 million minutes of conversation recorded across ten industry sectors. The firm’s call‑tracking and conversation‑analytics platform, developed by experienced analysts, captures this information reliably.
Notably, this is the first year Invoca has accumulated enough data to measure calls that stem from generative‑AI search tools.
What The Data Shows
Across all industries, roughly 56% of business calls are answered by a live person. The answer rate climbs with call length—about 65% for calls lasting more than 15 seconds and around 71% for those exceeding 30 seconds. Of the answered calls, approximately 38% qualify as leads, and close to 42% of those leads convert during the conversation.
ChatGPT‑referred calls perform better on the first metric—answer rate—but fall short of the overall average when it comes to lead qualification.
What The Report Doesn’t Say
Invoca has not disclosed the exact number of ChatGPT‑referred calls that underlie the 49% lead rate, only indicating that the total volume of generative‑AI‑driven calls is relatively small. Because the sample size is limited, rates derived from it may be less stable than those based on larger datasets, such as paid‑search traffic.
The report also omits any defined measurement window. While the methodology notes that the data reflect calls tracked on Invoca’s platform across ten industries and seven marketing channels, it does not specify the start or end dates of the period. Additionally, assistants such as Gemini, Claude, and Perplexity are omitted from the channel list; Invoca clarifies that this exclusion is a technical limitation, not a judgment on those tools, and that ChatGPT is the sole large language model currently producing measurable call volume in its dataset.
How Invoca Attributes The Calls
Invoca marks calls as ChatGPT‑referred, but the report does not detail the mechanics of that attribution—whether the caller clicked a tracked link from the assistant, dialed a number that was tagged in the system, or simply called after researching the business elsewhere. Consequently, the figures include only calls that can be directly linked to ChatGPT, excluding any subsequent calls that may have originated from an untracked source.
In a June piece, I noted that Similarweb data associated ChatGPT brand recommendations with a 2.5‑times higher likelihood of a site visit within a week. Most of that traffic showed up as branded search rather than a direct referral, which limits standard reporting. Adding calls to the mix introduces another attribution challenge, since the report does not explain the digital trail Invoca used to tie those calls back to ChatGPT.
Why This Matters
The data reveal that ChatGPT‑referred calls are about ten percentage points more likely to become qualified leads than calls from other channels tracked by Invoca. However, once a call is answered, the conversion rate mirrors the overall average, suggesting that the early‑stage advantage does not immediately boost closed deals.
I previously reported that Adobe observed a dramatic shift in AI‑referral conversion for U.S. retailers, moving from the lowest‑performing channel to a 42% improvement after a year. The prevailing view was that the research conducted inside the assistant primed the buyer before the call. Invoca’s findings align with that early‑stage insight: callers who have already compared options via ChatGPT may be nearer to a purchase decision.
Conversely, the later conversion outcome does not differ markedly. Although lead qualification is higher, the proportion of those leads that ultimately convert during the call remains close to the overall 42% average, indicating that the funnel narrows after the initial inquiry.
Looking Ahead
Invoca treats the finding as a development to watch rather than an immediate investment priority, citing the modest call volume as a limiting factor. The decisive metric is the number of calls generated, a figure not disclosed in the report. Looking forward, it will be important to see whether the 40% conversion rate holds steady or shifts as AI‑driven inquiries become more common. If AI‑referred callers keep qualifying at the top of the funnel while converting at average levels, the emphasis will move from volume growth to deeper insight into the conversation dynamics that drive actual sales.
Additionally, the report highlights that 64% of businesses fail to prompt callers to complete a purchase or schedule an appointment, underscoring a missed opportunity on the enterprise side.
