The test
Can Juma turn evidence into verified website changes?
Can I take a useful SEO or AEO recommendation from Juma, edit it, publish it to InterimCMO, and confirm the change without moving the work manually between tools?
I’ll test that process with new content, page titles, descriptions, schema, and internal links. The record will show what Juma handles, what I have to correct or build, and what the process costs.
Juma already documents a Webflow workflow connecting research, content creation, SEO and AEO edits, human review, and publishing. I have not independently tested that integration. This experiment asks whether I can do comparable work without choosing my website platform to fit the tool.
I’ll start with InterimCMO on ChatGPT Sites, then test the workflow on Replit. The goal is to carry the content, review rules, and test records with me, changing the publishing connection rather than rebuilding the whole process. Neither connection has been proven in this experiment yet.
Before the first test
Separate the starting point from Juma’s contribution.
The pre-experiment SEO and AEO baseline documents the website’s public launch, active analytics, and remaining search-measurement checks. I will save the approved site version and available measurements before Juma makes changes. Newsletter growth is the primary business outcome; publishing time, search visibility, and referral visits help explain it.
I will test both Juma’s own editorial suggestions and its refinement of a shortlist I bring to the session. The record will show what changed, what needed correction, and which ideas were worth pursuing.
Starting point
This is a return visit, not a first impression.
I have used Juma before. The test will separate what the product provides by default from the context, instructions, safeguards, integrations, and custom workflows I add as the operator. That distinction matters because a vendor's initial output is rarely the fully configured system.
The company
What Juma says it is
Juma, formerly Team-GPT, describes itself as a collaborative AI workspace and superagent for marketing teams and agencies. Its published materials describe shared Projects, persistent brand and project knowledge, prebuilt and custom Flows, agents, integrations, multi-model access, and Model Context Protocol connections. These are vendor-reported capabilities, not InterimCMO findings.
Juma uses credit-based pricing with unlimited seats. The test will record the plan, credits, overages, and support actually required rather than relying on a headline subscription price.
Sources: pricing, features, integrations, and MCP.
My take
Juma is a lightweight AI harness with purpose-built marketing capabilities.
A harness is the software around an AI model that manages its context, tools, and actions.
Viewed as a harness, Juma appears to bring several of these pieces together. My test is how well they work for a marketing team: sharing context, reviewing the work, and getting approved edits onto the live site correctly. The table separates its documented capabilities from what I still need to test.
| Harness layer | What Juma appears to provide | What the experiment must prove |
|---|---|---|
| Context and memory | Projects, brand rules, knowledge, instructions, and working history | Does Juma retain and apply the right context consistently? |
| Model orchestration | Automatic model selection and manual access to GPT, Claude, Gemini, and Perplexity | Do model choices improve the work without adding friction? |
| Workflow and agent building | Prebuilt Flows, custom Flows, and agents | Can the InterimCMO method be configured once and repeated reliably? |
| Marketing intelligence | SEO Pulse, integrations, web research, and connected data | Does the system produce evidence that is useful enough to act on? |
| Collaboration and control | Shared Projects, prompts, review, and history | Can a team collaborate consistently while people retain control? |
| Action and verification | Integrations, MCP, and documented Webflow publishing | Can an approved recommendation become a verified change on InterimCMO? |
| Learning and evaluation | Saved work, history, feedback, and usage reporting | Does the system improve over time, and can I see why? |
Included capability
SEO Pulse is the starting point.
Juma describes SEO Pulse as its native SEO toolkit, available without a separate connection. Its listed capabilities span keyword research, search intent, rankings, backlinks, technical SEO, competitive analysis, page parsing, audits, AI-search volume, LLM mentions, and cited domains and pages.
The first test is how much of the InterimCMO workflow SEO Pulse can support by itself. No additional paid SEO source will be added until a specific question cannot be answered credibly with SEO Pulse, the website, and Google Search Console.
The specialist benchmark
Use AirOps as the specialist benchmark.
AirOps is a purpose-built AI-search and content-engineering platform. It connects intelligence, content creation and review, CMS publishing, monitoring, and measurement. The comparison is useful because it shows what a specialized system makes possible.
The Juma experiment will not attempt to reproduce every AirOps feature. It will test whether a broader, configurable marketing harness can perform the parts of that workflow that matter to InterimCMO with fewer tools and less operational friction.
Ploy provides a second useful standard for the experience. Its reusable workflows connect research, structured page sections, metadata, schema, internal links, publishing, and monitoring. This experiment will not evaluate Ploy as a third product. It will test whether the Juma workflow can be saved as a reusable Method and deliver a similarly complete path from evidence to a verified website change without locking InterimCMO into one platform.
Cost will be evaluated rather than assumed. A lower subscription price only matters if setup, tuning, support, supporting tools, and operator time do not erase the savings.
Sources: AirOps platform, CMS integrations, and pricing; Ploy content-page workflow and Search Console workflow.
Minimum viable stack
Start with the fewest systems that can answer the questions.
- InterimCMO website GitHub-backed content, starting on ChatGPT Sites (OpenAI). Replit will be the next publishing target, not a requirement for the initial test.
- Juma The candidate harness and coordination layer.
- SEO Pulse The included source for SEO and GEO research.
- Google Workspace Docs for briefs, ideas, and working notes; Sheets for test records, timings, and costs; and Drive for organizing the supporting evidence.
- Google Search Console The planned source for Google Search performance. The verification DNS record is added; ownership verification, sitemap submission, and indexing checks remain open.
- Google Analytics Direct GA4 is active after explicit visitor consent. Page views and accepted newsletter and contact requests were verified in Realtime on September 14. Beehiiv confirms actual subscribers.
InterimCMO is public at interimcmo.com. Analytics remains opt-in worldwide. Search Console setup and a recorded observation period are still needed for the search baseline; the Juma experiment remains planned.
InterimCMO is a new site, so it does not yet have a stable traffic history. The initial test will focus on a small set of new editorial candidates: recommendation quality, human corrections, publishing reliability, time, and cost. From the September 14 public launch, I will follow discovery, search visibility, visits, and newsletter subscriptions over a stated period. Early growth will be an observation, not proof that Juma caused it.
Operating workflow
Connect the evidence to the website change.
Find an opportunity and the pages it affects.
Examine the evidence and assumptions.
Use AI-assisted writing to draft the page and internal-link changes.
Edit and approve the work.
Publish and verify the live change.
Measure the result and preserve the learning.
Questions
What the experiment needs to answer
- Recommendation qualityDoes Juma identify useful SEO and AEO opportunities, and what needs human correction?
- PublishingCan I review, edit, approve, publish, and verify titles, descriptions, schema, internal links, and page changes without manually moving work between tools?
- PortabilityCan I move from ChatGPT Sites to Replit without rebuilding the workflow or losing content, review rules, and test records? The goal is to support any AI-friendly website builder with a usable API, MCP server, or Git-based publishing path. I will report the connectors, custom work, time, and cost required for each platform tested, not assume universal compatibility.
- RepeatabilityCan the configured workflow become a reusable, portable Method, and what requires operator tuning, Juma support, or custom engineering?
- ValueDoes Juma, with SEO Pulse and any supporting tools, deliver enough of the AirOps capabilities that matter here to justify its total cost?
Some answers will be yes or no. Others will depend on the quality of the recommendation, the amount of operator correction required, the available budget, and the needs of this particular site.
Method
Six phases separate product capability from operator tuning.
Baseline
Record the current pages, metadata, internal links, visibility, referral traffic, and measurement limits.
Included system
Run Juma with its standard setup and SEO Pulse before adding custom instructions or another paid SEO product.
Operator-tuned
Add InterimCMO context, prompts, safeguards, custom Flows, and workflow changes, and record every meaningful correction.
Publishing
Start with an approved change to the GitHub-backed site, publish through ChatGPT Sites, and verify it live. Record any custom connection and manual steps required.
Support
Use Juma documentation first, then contact Juma support when blocked and record what support or workaround was required.
Repeatability
Run the workflow again to see whether it becomes faster, more accurate, and more useful. Then test publishing on Replit and record what carries over, what needs rebuilding, and the time and cost of switching.
True cost
Budget changes the alternatives, not the evidence.
The recommendation will consider three operating scenarios: a lean stack below $300 per month, a growing stack from $300 through $1,500 per month, and a larger budget above $1,500 that can support a specialized platform. The same evidence may lead to a different stack role in each scenario.
| Cost or tradeoff | What will be recorded |
|---|---|
| Subscription and usage | Plan price, credits or tasks, overages, and billing assumptions during the test. |
| Supporting tools | Any additional product required to complete the workflow credibly. |
| Setup and tuning | Time needed for access, context, custom instructions, Flows, integrations, and safeguards. |
| Operator time | Review, correction, approval, publishing, verification, and measurement effort. |
| Support | Documentation, vendor help, custom workarounds, and unresolved limitations. |
| Stack impact | Products the workflow can replace, products that remain necessary, and duplicated capabilities. |
Results
No results yet.
Get the findings in your inbox when they’re published.
This page currently records the plan. Findings, evidence, costs, configuration changes, support required, and limitations will be added as the experiment runs.
How the final recommendation will be stated
- Verdict Adopt, adopt with constraints, keep testing, or do not add.
- Best fit The team, budget, publishing system, and operating maturity the recommendation assumes.
- Stack role What Juma should own, what it can replace, and which supporting tools, specialist products, or custom connectors remain necessary.
- Support level Self-serve, vendor-supported, custom workaround, or unavailable.
Vendor involvement
Support is welcome. Editorial control is not for sale.
I’ll ask Juma support for help when I get stuck and report what it took to solve the problem. I’ll disclose any free access, sponsorship, or material assistance. Juma can correct product facts; I retain control of the findings and recommendation.