Can I try it for free?

Yes. When you sign up, you get 20 basic credits and 10 premium credits, valid for 6 months, plus 10 basic credits every day. Basic credits work for Existence checks (1 credit per reference); a Style check uses 2 premium credits per reference. Daily credits reset at 16:00 UTC and do not carry over, which is enough for occasional use. You can earn extra credits by completing your profile and by inviting friends who sign up, up to 100 credits in total from invitations. Each charge uses the day's basic credits first. Other credits are used in order of expiry, soonest first, with basic credits first when expiry dates match. Credits with no expiry date are used last, again basic credits first.

What files can I upload, and what can I export?

You can paste your references or upload a .docx (Word), .bib (BibTeX), .ris (EndNote, Zotero, and others), .enw, or .txt file. Reference lists export as .docx, .txt, .pdf, .bib, or .ris; check reports export as .pdf, .docx, or .md.

Can it check references from ChatGPT or other AI tools?

Yes. It doesn't matter where a reference came from. Paste references from ChatGPT, Claude, Gemini, or any other AI tool, and each one is compared against scholarly databases. AI-generated lists are where references to nonexistent papers turn up most often, so they are the ones most worth checking before you cite.

Does it use AI to check references?

Not for the verdicts: no AI is involved in deciding any verdict. AI only helps during parsing, splitting entries and identifying fields. The parsed fields are then checked against your original text field by field, and nothing beyond your original text is generated or rewritten.

Can I trust the results?

Every verdict is based on a field-by-field comparison with public scholarly databases such as Crossref, OpenAlex, and DataCite, covering authors, title, year, volume, issue, pages, DOI, and other fields. Each found or matched result links to its source, so you can check it yourself.

What does "Not found" mean?

It means the reference could not be found in the public databases we searched. That can happen for several reasons. The reference may be fabricated, its details may be wrong, or its fields may be mismatched. It may also simply not be indexed, as with recent preprints, some niche or regional journals, and books. We recommend searching for these references manually.

Which pack or pass should I buy?

It depends on how much you use it. A master's thesis usually cites at least 100 references, so a 300-credit pack covers checking and reformatting the whole list before your defense. For graduation season or a submission cycle, a pass is a better fit.

Will I be charged automatically?

No. Every purchase is a one-time payment, and nothing is charged automatically. A pass does not renew when it ends; to keep using the service, buy again.

Yes. Your research data is encrypted and stored securely, and we do not share your work with any third party. See our Privacy Policy.

Real. Complete. Correctly styled.
References, checked in one pass.

1.4B+scholarly records
9curated scholarly databases
41M+PubMed records

Check The verdict layer uses no generative AI, ruling out hallucinated or invented results; every field is backed by a record, every verdict by a source you can open.

460M+scholarly works
210K+journals
63M+full texts

Find Which source backs this sentence? Paste a paragraph, get relevant sources for each claim, and tick to cite.

Start free Find sources

10 free credits every day

01LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444. https://doi.org/10.1038/nature14539

Verified

Source

02Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of small-world networks. Nature, 393(6684), 440–442. https://doi.org/10.1038/30819

DOI discrepancy

Best match

DOI

− https://doi.org/10.1038/30819

+ https://doi.org/10.1038/30918

Accept and replace

03Thompson, R. J., & Patel, S. (2021). Automated verification of citation accuracy in academic manuscripts. Journal of Information Science, 47(6), 812–826.

Not found

Not found in public databases. This entry may be fabricated or not indexed. We recommend checking it manually.

Search manually:Google ScholarSemantic Scholar

LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. https://doi.org/10.1038/nature14539

Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, 313(5786), 504–509.

Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., & Riedmiller, M. (2013). Playing Atari with deep reinforcement learning. In NIPS Deep Learning Workshop.

Graduate School Office. (2016). Academic integrity workshop report [Unpublished internal report].

Upload a file (*.docx, *.bib, *.txt, *.ris, *.enw)

1/ 4 VerifiedPass rate 25%

Export

01LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444. https://doi.org/10.1038/nature14539

Verified

Deep learning

02Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, 313(5786), 504–509.

Volume/issue/pages differDOI missing

Best match

Pages

− 504–509

+ 504-507

DOI

+ https://doi.org/10.1126/science.1127647

03Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., & Riedmiller, M. (2013). Playing Atari with deep reinforcement learning. In NIPS Deep Learning Workshop.

Conference unconfirmed

Conference appears only in the preprint's author Comments note.

Playing Atari with Deep Reinforcement Learning

04Graduate School Office. (2016). Academic integrity workshop report [Unpublished internal report].

Not found

Not found in public databases. This entry may be fabricated or not indexed. We recommend checking it manually.

Search manually:Google ScholarSemantic Scholar

01Doudna, J. A., & Charpentier, E. (2014). The new frontier of genome engineering with CRISPR-Cas9. Science, 346(6213), 1258096. https://doi.org/10.1126/science.1258096

FoundComplete

Source

02Murray, C. J. (1994). Quantifying the burden of disease: The technical basis for disability-adjusted life years. Bulletin of the World Health Organization, 72(3), 429–45.

FoundComplete

Source

03Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., & Riedmiller, M. (2013). Playing Atari with deep reinforcement learning. In NIPS Deep Learning Workshop.

Conference unconfirmed

Source

ScienceAMERICAN ASSOCIATION FOR THE ADVANCEMENT OF SCIENCE

REVIEW SUMMARY

The new frontier of genome engineering with CRISPR-Cas9

Jennifer A. Doudna and Emmanuelle Charpentier

¹Howard Hughes Medical Institute, Department of Molecular and Cell Biology, University of California, Berkeley, CA, USA. ²Department of Chemistry, University of California, Berkeley, CA, USA. ³Helmholtz Centre for Infection Research, Braunschweig, Germany. ⁴Laboratory for Molecular Infection Medicine Sweden, Umeå University, Umeå, Sweden.

Science · 28 Nov 2014 · Vol 346, Issue 6213 · Article 1258096
DOI: 10.1126/science.1258096
eISSN 1095-9203 · Publisher: American Association for the Advancement of Science

Abstract

BACKGROUND: Technologies for making and manipulating DNA have enabled advances in biology ever since the discovery of the DNA double helix. But introducing site-specific modifications in the genomes of cells and organisms remained elusive. Early approaches relied on the principle of site-specific recognition of DNA sequences by oligonucleotides, small molecules, or self-splicing introns. More recently, the site-directed zinc finger nucleases (ZFNs) and TAL effector nucleases (TALENs) using the principles of DNA-protein recognition were developed. However, difficulties of protein design, synthesis, and validation remained a barrier to widespread adoption of these engineered nucleases for routine use.

ADVANCES: The field of biology is now experiencing a transformative phase with the advent of facile genome engineering in animals and plants using RNA-programmable CRISPR-Cas9. The CRISPR-Cas9 technology originates from type II CRISPR-Cas systems, which provide bacteria with adaptive immunity to viruses and plasmids. The CRISPR-associated protein Cas9 is an endonuclease that uses a guide sequence within an RNA duplex, tracrRNA:crRNA, to form base pairs with DNA target sequences, enabling Cas9 to introduce a site-specific double-strand break in the DNA. The dual tracrRNA:crRNA was engineered as a single guide RNA (sgRNA) that retains two critical features: a sequence at the 5′ side that determines the DNA target site by Watson-Crick base-pairing and a duplex RNA structure at the 3′ side that binds to Cas9. This finding created a simple two-component system in which changes in the guide sequence of the sgRNA program Cas9 to target any DNA sequence of interest. The simplicity of CRISPR-Cas9 programming, together with a unique DNA cleaving mechanism, the capacity for multiplexed target recognition, and the existence of many natural type II CRISPR-Cas system variants, has enabled remarkable developments using this cost-effective and easy-to-use technology to precisely and efficiently target, edit, modify, regulate, and mark genomic loci of a wide array of cells and organisms.

OUTLOOK: CRISPR-Cas9 has triggered a revolution in which laboratories around the world are using the technology for innovative applications in biology. This Review illustrates the power of the technology to systematically analyze gene functions in mammalian cells, study genomic rearrangements and the progression of cancers or other diseases, and potentially correct genetic mutations responsible for inherited disorders. CRISPR-Cas9 is having a major impact on functional genomics conducted in experimental systems. Its application in genome-wide studies will enable large-scale screening for drug targets and other phenotypes and will facilitate the generation of engineered animal models that will benefit pharmacological studies and the understanding of human diseases. CRISPR-Cas9 applications in plants and fungi also promise to change the pace and course of agricultural research. Future research directions to improve the technology will include engineering or identifying smaller Cas9 variants with distinct specificity that may be more amenable to delivery in human cells. Understanding the homology-directed repair mechanisms that follow Cas9-mediated DNA cleavage will enhance insertion of new or corrected sequences into genomes. The development of specific methods for efficient and safe delivery of Cas9 and its guide RNAs to cells and tissues will also be critical for applications of the technology in human gene therapy.

AbstractFull textFiguresReferencesInfo & Metrics

01Doudna, J. A., & Charpentier, E. (2014). The new frontier of genome engineering with CRISPR-Cas9. Science, 346(6213), 1258096.

02Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, 313(5786), 504-507. https://doi.org/10.1126/science.1127647

03He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 770-778).

04Pearl J. Causality: models, reasoning, and inference. 2nd ed. Cambridge: Cambridge University Press; 2009.

05Watson, J. D., & Crick, F. H. C. (1953). Molecular structure of nucleic acids: A structure for deoxyribose nucleic acid. Nature, 171(4356), 737-738. https://doi.org/10.1038/171737a0

APA 7

Doudna, J. A., & Charpentier, E. (2014). The new frontier of genome engineering with CRISPR-Cas9. Science, 346(6213), 1258096.

Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, 313(5786), 504–507. https://doi.org/10.1126/science.1127647

He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 770–778).

Pearl, J. (2009). Causality: Models, reasoning, and inference (2nd ed.). Cambridge University Press. https://doi.org/10.1017/cbo9780511803161

Watson, J. D., & Crick, F. H. C. (1953). Molecular structure of nucleic acids: A structure for deoxyribose nucleic acid. Nature, 171(4356), 737–738. https://doi.org/10.1038/171737a0

Type

File

Report.pdf

PerfectRefs

Generated 2026-07-13 13:00 · Style: APA 7

Reference Verification Report

This check covers 10 entries: Needs correction 3Needs checking 1Not found 2Verified 4

Entry 1 · Verified

LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444. https://doi.org/10.1038/nature14539

Source: https://doi.org/10.1038/nature14539

Entry 2 · Needs correction

Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of 'small-world' networks. Nature, 393(6684), 440–442.

A matching record was found, but there are discrepancies in DOI:

DOI: "10.1038/30918".

Correction: Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of 'small-world' networks. Nature, 393(6684), 440–442. https://doi.org/10.1038/30918

Source: https://doi.org/10.1038/30918

Entry 3 · Needs correction

Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, 313(5786), 504–509. https://doi.org/10.1126/science.1127647

A matching record was found, but there are discrepancies in Pages:

Pages: "504-507".

Correction: Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, 313(5786), 504–507. https://doi.org/10.1126/science.1127647

Source: https://doi.org/10.1126/science.1127647

Entry 4 · Verified

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017). Attention is all you need. In Advances in Neural Information Processing Systems 30 (pp. 5998–6008).

Source: https://arxiv.org/abs/1706.03762v7 · https://www.semanticscholar.org/paper/204e3073870fae3d05bcbc2f6a8e263d9b72e776

Entry 5 · Needs checking

Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., & Riedmiller, M. (2013). Playing Atari with deep reinforcement learning. In NIPS Deep Learning Workshop.

Journal: "NIPS Deep Learning Workshop", Journal appears only in the author's note.

Source: https://arxiv.org/abs/1312.5602v1

Entry 6 · Verified

Doudna, J. A., & Charpentier, E. (2014). The new frontier of genome engineering with CRISPR-Cas9. Science, 346(6213), 1258096. https://doi.org/10.1126/science.1258096

Source: https://doi.org/10.1126/science.1258096

Entry 7 · Verified

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291. https://doi.org/10.2307/1914185

Source: https://doi.org/10.2307/1914185

Entry 8 · Needs correction

Walters, W. H., & Wilder, E. I. (2022). Fabrication and errors in the bibliographic citations generated by ChatGPT. Scientific Reports, 13, Article 14045. https://doi.org/10.1038/s41598-023-41032-5

A matching record was found, but there are discrepancies in Year, Issue:

Year: "2023";

Issue: "1".

Correction: Walters, W. H., & Wilder, E. I. (2023). Fabrication and errors in the bibliographic citations generated by ChatGPT. Scientific Reports, 13(1), Article 14045. https://doi.org/10.1038/s41598-023-41032-5

Source: https://doi.org/10.1038/s41598-023-41032-5

Entry 9 · Not found

Thompson, R. J., & Patel, S. (2021). Automated verification of citation accuracy in academic manuscripts. Journal of Information Science, 47(6), 812–826.

Not found in public databases.

Entry 10 · Not found

Graduate School Office. (2016). Academic integrity workshop report [Unpublished internal report].

Not found in public databases.

Note: "Not found" means not found in public databases.

How we check every reference

PerfectRefs splits your reference list into individual references and looks up each one in several established scholarly databases. If the usual databases have no match, it automatically searches more. It compares authors, title, source, year, volume, issue, and other fields, and decision rules we built ourselves produce the verdict. A style check extracts the fields your chosen citation style needs, with AI helping to parse complex references. Those fields are then compared one by one with the database record. Wherever a field differs, you see the correct value from the record and can accept it in one click. Every step in the diagram is a step each reference really goes through.

Every verdict has a source.

Each field is compared with public scholarly databases: authors, title, year, volume, issue, pages, DOI and more. Open the source behind any verdict to check it yourself.

CrossrefOpenAlexPubMedarXivSemantic ScholarDataCitedoi.org

+ 2 Chinese-language databases

Try it first, then buy what you need

See full pricing

Free

10 credits a day, reset daily.

Most flexible

Credit packs · for a set amount

Passes · for a set period

Put your references through this check first.

Start checking

Before your advisor, committee or journal sees it

Real. Complete. Correctly styled.References, checked in one pass.

The new frontier of genome engineering with CRISPR-Cas9

How we check every reference

Every verdict has a source.

Try it first, then buy what you need

Put your references through this check first.

Frequently asked questions

Can I try it for free?+

What files can I upload, and what can I export?+

Can it check references from ChatGPT or other AI tools?+

Does it use AI to check references?+

Can I trust the results?+

What does "Not found" mean?+

Which pack or pass should I buy?+

Will I be charged automatically?+

Is my data safe?+

The new frontier of genome engineering with CRISPR-Cas9

Real. Complete. Correctly styled.
References, checked in one pass.