Know if you can trust the research.
A clear-eyed reading of every paper, claim, and citation — with the deductions written out beside the score.
No credit card · 10 free readings per month
A Review of Climate Sensitivity Estimates
Yusifova, A. · Asadov, R. · 2024
We revisit equilibrium climate sensitivity using the latest instrumental temperature record alongside the published paleoclimate constraints. Compared with the Sherwood et al. synthesis, our estimate narrows the lower tail by an order of magnitude. We attribute the shift to two factors. First, the corrected sea-surface temperature reanalysis removes a systematic cooling bias[7]. Second, the revised aerosol forcing budget brings the short-wave cloud feedback into the range reported by CESM2-WACCM and HadGEM3. The combined effect repositions the central estimate at 3.1 K with a 90% range of 2.4 to 4.0 K, consistent with the IPCC AR6 likely range.
Excerpt — full report available after analysis.
Verified — Nature 2024 retraction archive.
Reference 7 doesn't support the claim.
Paragraphs 3, 5, 11 — likelihood >0.7.
Built for 51 universities and 222,809 students across Azerbaijan and the Eastern Partnership region.
ADA · BSU · Khazar
Retractions detected each year
1Retraction Watch · 2024
Students using AI writing tools
2HEPI Survey · 2024
Mean cost of a retracted paper
3eLife · 2024
Time to a full reading
4ArtiProof · telemetry
Numbers are recorded findings, not estimates. Each citation is preserved in the published report so a reader can re-trace the chain themselves.
Three exhibits. One scoring rubric.
Pick an exhibit to read the report a real submission would produce. The data is canned for these three; the format — claim verdicts, AI probability, citation status, credibility score — is what every reading carries.
Each engine returns evidence, not a verdict.
Source checks and researcher lookups answer different questions. These examples illustrate the tools; inspect the actual findings and coverage in each analysis.
Inspect the claims selected for review.
Selected claims are compared with existing fact-checks and analyzed by a language model. Inspect the verdict, explanation and available links; a result may remain unverified.
Method ·Verifies claims against the Google Fact Check API
Fig. 1 · synthesized output
Read the verdicts, not a percentage.
True, false, misleading, partially true, or unverified. Available source links help you investigate a verdict; model-generated explanations are not independently verified evidence.
Method ·Up to 15 claims per submission
matches NOAA 2024 record
study retracted Apr 2023
context lost in revision
Fig. 2 · synthesized output
Inspect paragraph-level style signals.
The model estimates AI likelihood from a sample of paragraphs, with a heuristic fallback when unavailable. These style signals can be wrong and do not establish authorship.
Method ·Analyzes up to 20 paragraphs per document
Fig. 3 · synthesized output
Explore frequent citing authors.
Researcher analysis samples incoming citation relationships. A concentration of citing authors can guide further review; it does not establish reciprocal citation rings or manipulation.
Method ·Incoming citation samples, not proof of a ring
Fig. 4 · synthesized output
Compare reported and estimated metrics.
A separate researcher lookup collects publication metrics and estimates adjustments from sampled citation data. Adjusted h-index values are estimates, not a paper-by-paper recomputation.
Method ·Reported metrics alongside estimates
Illustrative reported and adjusted metrics. Estimates depend on the sampled citation data and require review.
Fig. 5 · synthesized output
Native Azerbaijani and Russian. Local compliance. Pricing that works for a regional university budget — not a US R1.
Know what was checked. Know what needs review.
Source checks and researcher lookups are separate workflows. Use the findings to guide a review, with the coverage and limitations alongside the result.
| Capability | Current output | Limitations |
|---|---|---|
| Source checks | Extracted claims, verdicts and source links when available. | Up to 15 claims from the first 15,000 characters. Similar fact-check matches need human confirmation. |
| AI style estimates | Paragraph-level signals with a heuristic fallback. | Samples up to 20 paragraphs. Estimates are not proof of authorship or misconduct. |
| Citation checks | Link accessibility and retraction signals for links extracted from web articles. | PDF and pasted-text bibliographies are not yet parsed into a reference inventory. Missing retraction data does not establish a clean record. |
| Researcher analysis | A separate lookup for publication metrics and citation patterns. | Frequent incoming citations do not establish reciprocal citation rings or manipulation. Results require review. |
| Reports | Downloadable reports and revocable share links. | Reports summarize the analysis performed. They do not certify a source or researcher. |
| Plagiarism and LMS integrations | Not available in the current product. | Corpus-based plagiarism checking and native Moodle/Canvas plugins are not implemented. |
Free for the first ten readings. $20 if you keep going.
Start on Free. Pro prices are proposed; registration does not activate a paid subscription. Institutional capabilities and terms require a scoped agreement.
Free
— for the curious
- 10 source readings per month
- Paragraph-level AI style estimates
- Credibility scoring
- 5 researcher analyses per month
- 5 PDF reports per month
- Citation and self-citation analysis
- API access
No credit card required
Pro
— the working researcher
- 100 source readings per month
- 50 researcher analyses per month
- Citation and self-citation analysis
- Paragraph-level AI style estimates
- Claim review with source links when available
- 50 PDF reports per month
- Public API subscriptions
Registration starts on Free; paid activation is separate
Team
— a research desk that ships together
starter · up to 10 seats
- Everything in Pro for every seat
- Up to 10 seats (Starter) · 25 seats (Pro · $249/mo)
- Shared watchlists + team activity feed
- Role-based admin / member controls
- Per-seat usage dashboard
- Pooled report exports + branded PDFs
- Audit log of every team action
Free for 14 days · no card required
Institution
— the university press
- Discuss a scoped institutional pilot
- Agree seats, usage and retention before rollout
- Department dashboards (planned)
- Native LMS plugins (planned)
- Institutional SSO (planned)
- On-premise deployment (planned)
- Dedicated support terms by agreement
Scope and commercial terms by agreement
Submitted text, uploaded PDFs and analysis findings are stored to support your history and reports. Deletion follows the service's retention process; content is not automatically discarded when analysis finishes.
Begin a reading in ninety seconds.
Free for the first ten readings. No credit card. No commitment. Bring a paper, a URL, or a paragraph.
No credit card · 10 readings / month · Cancel anytime
Built for researchers across ADA University, Baku State University, and Khazar University.