Genspark is one of the more interesting agent products of the 2024-2026 cycle. Founded by Eric Jing and Kay Zhu (both ex-Baidu), it sits between search engine and chat assistant, with a signature output called a Sparkpage: a synthesised, source-cited research page generated in minutes. Through 2025 it added Calling Agent, Slides Agent, Sheets Agent, and a Super Agent capability that broadens the product into delegated task execution (Genspark, 2026).

I respect what Genspark has built. It is also a different category from Gravity, and one of the more common buyer mistakes is to compare them on a single grid as if they were direct substitutes. They are not. Both produce real value; the value lives at different surfaces.

What Genspark actually is in 2026

Genspark launched in 2024 as an AI search experience. The first Sparkpage demos showed a single query producing a structured, cited research page faster than traditional search and richer than a single chat response. The team raised significant funding through 2024-2025 and shipped Calling Agent (an agent that places phone calls on the user's behalf), Slides Agent (generates structured presentation decks), Sheets Agent, and Super Agent (a meta-agent orchestrating the specialised ones).

The Sparkpage

The Sparkpage is the artifact most associated with Genspark. Submit a query, get back a structured page with sections, cited sources, comparisons, and synthesised analysis. The output is meant to be read, copied, or repurposed; it is not meant to live and update as a system of record.

The specialised agents

Calling, Slides, Sheets, Super. Each is excellent at its job. Calling Agent in particular has impressed reviewers who tested it on restaurant reservations and outbound prospecting. The shared property: each agent produces a deliverable per invocation, not a deployed long-running service.

The research-output category (and why it is bigger than people think)

Knowledge workers spend a meaningful share of their day on research and synthesis. A 2024 McKinsey study estimated that 60-70% of knowledge worker tasks have an automatable component, with research and synthesis among the highest-volume buckets (McKinsey, 2024). Genspark's category is real, valuable, and large.

The research-output category is defined by three properties. First, the unit of work is the deliverable, not the deployment. Second, the user reads the output before acting on it. Third, recurrence is rare; the same exact research is not run daily.

What Gravity does differently

Gravity is built for the opposite shape of work. The unit of work is the deployed agent. The user does not read each output; they trust the system to act. Recurrence is the default; the agent runs every day, every Monday, on every new lead, forever. For the broader framing see describe outcome, not workflow and AI agent vs chatbot vs assistant.

The implications shape every Gravity design choice. Integrations have to write back to systems of record. Monitoring has to surface across thousands of runs. Failure modes have to recover automatically because nobody is reading each output. Cost has to be predictable because the work runs unattended.

Research agent vs operational agent: the framework

DimensionResearch agent (Genspark)Operational agent (Gravity)
Unit of workDeliverable per queryDeployed long-running agent
Output destinationUser reads itSystem of record updates
RecurrenceOne-shot or rareDaily / weekly / on-event
Integration breadthWeb sources, search, light SaaSCRM, billing, ticketing, comms
Trust modelUser reviews each outputAudit logs, recovery, monitoring
Cost dynamicsPer queryPer outcome, predictable
Best buyerAnalyst, consultant, founderOperator, ops lead, founder

Capability comparison

One axis where I would not pretend Gravity wins: speed to a single research deliverable. If you want a five-page comparison of three competitors with sources, in five minutes, Genspark is faster than any operational platform. The flip side is that Gravity is faster at producing daily outcomes against integrated systems.

Output formats

Genspark produces pages, slides, sheets, calls. Gravity produces state changes in systems of record, plus optional summaries. Different formats serve different jobs.

Integration depth

Genspark's strength is web-scale source retrieval. Gravity's strength is integration depth with business SaaS. Both are legitimate. Most teams need a mix.

Where Genspark is the right tool

Three categories.

Sales preparation

"Tell me everything notable about this prospect's company in the last 60 days." Sparkpages are excellent here. The output is read once, before a meeting, and discarded.

Market scans and competitor analysis

"Compare these five competitors on pricing, positioning, and recent product launches." One-shot research with a clean deliverable.

One-off outreach calls

Genspark Calling Agent can place a single outbound call (restaurant booking, vendor follow-up) better than most alternatives. The unit of work is the call, not the relationship.

Where Gravity wins

Three opposite categories.

Recurring lead follow-up

"Follow up with every new lead within 10 minutes, with personalised context, forever." Recurring, integrated, audit-required. Operational territory. See AI agent orchestration for the deeper view.

System-of-record updates

"Reconcile every Stripe payment against the CRM contact and the accounting system." The output is not a page to read; it is a state change across systems. Operational platforms are built for this. See AI agent deployment models.

Multi-step real-world execution

"When a customer cancels, archive their workspace, refund pro-rated, notify CS, schedule the win-back email in 30 days." Multi-step, integrated, recurring. The operational category. See AI agent failure modes and three startups, three shutdowns.

Frequently asked questions

What is Genspark and what does it do?

Genspark is an AI research and agent product from ex-Baidu founders, launched in 2024 and expanded through 2025 with Calling Agent, Slides Agent, Sheets Agent, and Super Agent capabilities. The signature output is the Sparkpage: an AI-synthesised research page that pulls from many sources into a single artifact. The category is research and one-shot deliverable, not deployed recurring ops.

Are Genspark and Gravity competitors?

Adjacent, not direct. Genspark sits in the research-output category (one-shot synthesis, deliverable in minutes). Gravity sits in the operational-agent category (deployed daily work, ongoing integration with systems of record). Some tasks could be done in either, but the typical use case maps cleanly to one or the other.

What is a Sparkpage and when should I use one?

A Sparkpage is Genspark's synthesised research page: a single document pulling structured findings from many web sources. Use it for sales prep, market scans, competitor analysis, and ad hoc fact-finding. Do not use it for recurring operations where the output has to update systems of record or trigger downstream actions.

Can Genspark Super Agent do operational work?

Genspark's Super Agent and its specialised agents (Calling, Slides, Sheets) extend the product beyond pure research into task execution. They handle one-shot delegated tasks well. For recurring deployed work with integration into business systems and per-outcome economics, a focused operational platform fits better.

When is Gravity a better fit than Genspark?

When the work is recurring rather than one-off; when the output needs to write back to a system of record (CRM, billing, ticketing); when integration breadth across business SaaS matters more than research depth; and when per-outcome economics and audit trails are part of the procurement story.

Three takeaways before you close this tab

Sources