Soku

Soku is an always-on AI marketing agent designed to connect the analysis and execution parts of growth work. Its current web product spans paid campaigns, creative generation, cross-channel reporting, anomaly monitoring, SEO/social workflows and analytics across connected accounts including Meta Ads, Google Ads, TikTok Ads, GA4, Shopify and many additional integrations.¹²

The key control model is graduated autonomy. Teams can begin with analysis-only access, route write-capable changes through approval, or enable automation rules for routine actions. Supported actions are recorded in an Activity log, while the product can also be used through a CLI and a separate Connect/MCP surface from tools such as Claude, ChatGPT, Cursor and Codex.¹³

Soku is developing quickly. Its August 2026 changelog documents reliability fixes for long-running conversations, account connections and campaign validation, as well as expanded integrations, ChatGPT Ads workflows and a dedicated Connect console.⁴

Soku uses credit-based pricing. The current page lists Free at $0 with 2,000 starting credits; Creator at $15/month on annual billing, with the page's pricing explainer describing $19 monthly; Starter at $39/month on annual billing with 240,000 credits/year; Pro at $199/month on annual billing with 2.4 million credits/year; and custom Enterprise with broader security and success support.⁵

Paid Starter and Pro organizations can buy one-time credit packs at $3 per 1,000 credits, with purchases from 10,000 to 100,000 credits. Credit consumption varies by workflow depth and model/provider use, so the subscription headline alone does not determine operating cost.⁵

Current independent evidence is sparse. G2 has an exact Soku AI profile but currently shows zero reviews, and its product description still characterizes Soku primarily as an ad-research engine rather than the broader autonomous marketing-agent surface now documented by Soku itself.⁶ It is therefore useful as product-identity evidence but not as current capability or quality evidence.

The current first-party product includes explicit safeguards that matter for a system with ad-account write access: users can begin read-only where supported, write-capable workflows must be enabled separately, approvals can gate risky actions, and recorded activity can be reviewed afterward.¹ The changelog also provides unusually detailed disclosure of reliability defects and fixes, which improves transparency but is not an independent reliability benchmark.⁴

No reviewed source supplies a mature verified-user corpus or controlled independent benchmark of campaign outcomes attributable to Soku. Vendor ROI and time-saved calculators should therefore be treated as planning models and first-party pilot claims rather than independently established returns.⁵

Soku is best for performance teams and agencies that already operate across multiple paid-media and analytics systems and spend substantial time assembling reports, spotting anomalies and turning recommendations into account changes. Its cross-brand and client-reporting surfaces make the agency use case particularly strong.¹

It is less suitable for teams that need a mature independent evidence base before granting software write access to production ad accounts, or whose marketing work is mainly high-level brand strategy rather than recurring measurable execution.