AgentSEO Research
Product claims should come with a method.
Inspect what we measured, what the result does not prove, and which evidence path matches the workflow you are evaluating.
Research navigator
Choose the proof you need before you wire the API.
The right evidence depends on the job. Pick your operating context to see what AgentSEO has published, what remains unproven, and the shortest responsible verification path.
Developer path
Will this output fit inside an agent loop without a payload-cleanup layer?
Published evidence
A reproducible three-query benchmark compares provider task-result bytes with AgentSEO core and enriched output.
Evidence gap
The sample is directional. It does not establish token savings, latency performance, or results for every endpoint.
Verification path
- 01
Inspect the exact queries, byte measurement, and exclusions.
- 02
Compare the normalized and enriched response shapes.
- 03
Run one representative request against your own workflow.
Growth engineer path
Can this become a repeatable operating loop instead of another dashboard export?
Published evidence
The current benchmark tests payload shape and publishes machine-readable aggregates with its method and limits.
Evidence gap
A public workflow-usefulness study is not published yet. Treat the workflow bundles as implementation guides, not outcome proof.
Verification path
- 01
Pick one monitored job: launch, refresh, AI visibility, or competitor gap.
- 02
Define the decision the output must support before you automate it.
- 03
Measure review time, follow-up prompts, and useful decisions in your environment.
Agency path
Can I predict the operating cost and review the evidence before client delivery?
Published evidence
Public pricing explains credits and hard caps; the research page shows how AgentSEO labels small samples and quantitative limits.
Evidence gap
No public multi-client throughput study or agency outcome benchmark is available yet. Do not infer one from payload reduction.
Verification path
- 01
Map one client deliverable to its endpoint and expected run frequency.
- 02
Estimate credits with the published pricing model.
- 03
Keep a human review gate until output quality is stable across accounts.
Search payload size benchmark
A three-query live comparison of provider task-result payloads, AgentSEO normalized JSON, and the enriched output supplied to an agent workflow.
Read the benchmark- Sample
- 3 queries
- Core reduction
- 88.6%
- Enriched reduction
- 72.6%
- Measured
- UTF-8 bytes
Runnable scope
Each benchmark identifies inputs, environment, sample size, and exactly what was measured.
Public data
Machine-readable aggregates are published when disclosure does not expose customer or provider-sensitive data.
Visible limits
Small samples stay labeled as small samples. Directional findings are not promoted into universal claims.
What We Are Testing Next
The research surface is moving from payload size to workflow fit.
Payload reduction still matters, but the current product question is broader: can an agent turn SEO data into a safe, repeatable workflow decision without wasting credits, tokens, or operator time?
Workflow usefulness
Measure whether bundled workflows help builders move from one endpoint call to a real SEO job: launch, refresh, AI visibility, competitor gap, or backlink opportunity work.
Credit-cost predictability
Track how deterministic 2-credit workflows, paid provider calls, and usage-scaled monitoring behave in real jobs so pricing stays transparent and margin-safe.
Agent-ready output quality
Validate that JSON plus markdown_summary outputs reduce follow-up prompting and make downstream Claude Code, MCP, and automation workflows easier to operate.