How to Check for Hallucinations in AI Text: Causes, Examples, Risks, and Checklist

Key Takeaways:

• AI hallucinations are confident-sounding statements, citations, or numbers that are not grounded in real sources, and they occur in every academic field.
• Researchers and students must verify every citation, quote, and statistic against a primary source before using AI-generated text in their work.
• A professional editor can improve clarity, tone, and structure while cross-checking citations, but only the author can confirm that quotes, data, and code are accurate.
• ESL authors and first-time authors face a higher risk of missing hallucinations because fluent, well-structured prose can hide factual errors.

Glossary of Key Terms

Term	Definition
Hallucination	A statement, citation, or data point generated by an AI system that sounds plausible but is not based on real facts or sources.
Fabricated citation	A reference that looks legitimate, complete with author names and journal titles, but does not exist or does not say what the text claims.
Confabulation	A term borrowed from psychology, used to describe an AI system filling gaps in knowledge with invented but coherent-sounding details.
Grounding	The process of linking AI-generated statements to verifiable, real-world sources or data.
Retrieval-augmented generation (RAG)	A technique that lets an AI system pull information from external documents before generating text, which can reduce hallucinations.
Fact-checking	The process of confirming that a claim, number, or quote matches an original, reliable source.
Source verification	Checking that a cited source exists, is accessible, and actually supports the claim attached to it.
Plagiarism check	A scan that compares text against published work to detect copied or insufficiently paraphrased content.

What Is a Hallucination in AI Text?

A hallucination is a statement, citation, or number that an AI tool generates with confidence but that has no basis in real facts or verifiable sources.

Large language models predict the next likely word based on patterns in training data. They do not check facts against a live database unless connected to a search or retrieval tool. When a model lacks reliable information on a topic, it often produces a plausible-sounding answer instead of admitting uncertainty. This is why AI text can include invented authors, incorrect dates, or numbers that don’t actually come from your dataset.

Hallucinations differ from simple errors. A typo is a mistake in form, while a hallucination is a mistake in substance, where the AI invents content that did not exist before. For researchers and students, this distinction matters because hallucinated content can look identical in style and tone to accurate content.

Some AI tools reduce hallucinations through grounding or retrieval-augmented generation, which lets the model pull information from real documents before writing a response. Even with these techniques, errors still occur, especially on niche topics, recent events, or subjects with limited training data. No current AI tool guarantees a hallucination-free output, so manual verification remains necessary for any academic or professional document.

Why Do AI Models Hallucinate?

AI models hallucinate because they are built to predict likely wording, not to retrieve verified facts, so when reliable information is missing they produce a confident guess instead of an admission of uncertainty.

Understanding the cause matters because each cause has a different fix. A hallucination that comes from a knowledge cutoff is solved by supplying the source. One that comes from a leading prompt is solved by rewriting the prompt.

Why fluent language doesn’t make AI accurate

• Fluency and accuracy are produced by two different processes inside the model, so polished phrasing carries no signal about whether a claim is true.
• A fabricated citation is generated with the same confidence, formatting, and tone as a real one, which is why hallucinations survive proofreading.
• Readers slow down at awkward sentences and speed up at smooth ones, so the most fluent passages tend to receive the least scrutiny.
• This is the reason verification has to be a separate, deliberate pass rather than something done while reading the draft.

Why This Matters for Researchers and Students

Academic and professional writing depends on accuracy. A single fabricated citation or incorrect statistic can undermine an entire paper, delay publication, or damage a student’s academic record. Journals and universities increasingly run AI-detection and fact-checking software, so hallucinated content is more likely to be caught than in the past. One of the most popular preprint services, arXiv.org, recently imposed a one-year ban on authors who submitted papers with fabricated references.

Beyond detection risk, there is a research integrity issue. Citing a source that does not exist misleads readers who may try to locate it. Presenting an invented statistic as real can influence how other scholars interpret evidence. Checking for hallucinations protects both the author’s credibility and the wider body of knowledge that other researchers rely on.

There is also a time cost to consider. Correcting a hallucinated citation after a paper is under review is far more disruptive than catching it before submission. Reviewers and editors who find even 1 fabricated reference may question the reliability of the entire manuscript, which can slow down or end a review process that took months to reach.

Hallucination in AI Models: Do Some Tools Hallucinate Less Than Others?

Hallucination in AI models varies by how the tool gets its information rather than by brand, and tools that read real documents before answering fabricate fewer citations than tools answering from internal memory.

Why grounding reduces hallucination without removing it

• Retrieval confirms that a document exists, but it does not confirm that the model read the relevant section correctly.
• A model can cite a real, high-quality paper and still attribute a claim to it that the paper never makes.
• Retrieved sources can themselves be wrong, outdated, predatory, or retracted, and the model rarely flags this.
• Long research-style outputs feel authoritative precisely because they carry many citations, so they tend to be spot checked rather than fully checked.
• Every citation in a grounded output still has to pass the citation verification workflow described earlier in this article.

Do vendor-provided hallucination rates mean anything for academic writing?

Just because a vendor has published a very low hallucination rate, it doesn’t mean that the AI tool will not hallucinate when you use it, for the following reasons:

• Rates are measured on a specific model version, and versions change frequently without notice.
• Results depend heavily on the field tested, since well-documented topics produce far fewer fabrications than niche, complex, or emerging topics.
• Prompt wording changes outcomes substantially, so two studies of the same tool can report very different figures.
• A low reported rate still means there could be some fabricated content in your reference list, which is why no rate justifies skipping verification.

What Are the Common Types of AI Hallucinations?

The 4 most common types are factual errors, fabricated citations, incorrect numbers or statistics, and logical claims that do not follow from the evidence presented.

Factual hallucination

The AI states something false as if it were established fact, such as the wrong founder of an organization or an incorrect date for a historical event.

Citation hallucination

The AI invents an author, title, journal, or page number, or attaches a real author’s name to a paper they never wrote.

Numerical hallucination

The AI reports a statistic, percentage, or measurement that was never published or that contradicts the original source.

Logical hallucination

The AI draws a conclusion that sounds reasonable but does not actually follow from the data or argument presented earlier in the text.

These 4 categories often overlap in a single document. A fabricated citation, for example, frequently supports a numerical hallucination, since the invented source is used to justify an invented statistic. Recognizing which type of error you are looking at helps you decide which verification method to use, whether that is a database search, a recalculation, or a rereading of the original argument.

ChatGPT Hallucination: What It Looks Like in Academic Work

A ChatGPT hallucination in academic writing usually appears as a correctly formatted reference built from real components that were never combined in a real publication.

The most common pattern is a composite citation. Individual elements are recognizable, which is what makes the reference survive a quick glance.

Anatomy of a fabricated reference

Why ChatGPT identifiers are especially deceptive

• DOIs and PMIDs are generated in the correct format, so they pass a visual check and often pass a reference manager import.
• A DOI prefix belonging to a genuine publisher makes the identifier look verified before anyone tests it.
• Some fabricated DOIs resolve to a real but unrelated paper, which can be more damaging than a dead link because the citation appears to check out.
• Reference managers will store a fabricated entry without throwing up an error message, so a clean bibliography is not evidence of a verified one.

What changes with browsing and file upload, and what does not

ChatGPT users often feel that enabling browsing and uploading their own files can reduce the number of hallucinations. But this is not a foolproof method.

Where ChatGPT is reliable and where it is not

Does ChatGPT admit when it is wrong?

• Asking the tool whether its citations are real is not a verification method, because the same process that generated the citation generates the confirmation.
• Under mild pushback, models often reverse a correct answer, which means agreement and retraction are both unreliable signals.
• Asking for a confidence rating produces a number that reads precisely but is not calibrated against anything.
• The only reliable check remains opening the source yourself, which is the workflow set out earlier in this article.

Field-Specific Examples of AI Hallucinations

Hallucinations appear differently depending on the subject. The examples below show how the same underlying problem, invented content presented with confidence, takes different forms across 6 fields.

Medicine and Health Sciences

An AI summary of a drug trial might state that a medication reduced symptoms by 45% when the original study reported 28%. It might also cite a clinical guideline that was updated or retracted, without noting the change. In health writing, an unverified number can affect patient safety, not just academic accuracy.

Law and Policy

AI tools have been documented citing court cases that do not exist, complete with realistic-sounding case names and docket numbers. A law student using AI for a brief might receive a citation to a case that sounds authoritative but cannot be found in any legal database. Courts have sanctioned attorneys for filing such fabricated citations.

History and Social Science

An AI might describe a treaty as signed in 1919 when the correct year is 1920, or attribute a policy to the wrong administration. It may also blend details from 2 separate historical events into 1 account, creating a hybrid narrative that never actually happened.

Computer Science and Engineering

AI-generated text on a technical topic may reference a software library version that does not exist or describe a function that behaves differently from the real documentation. A student citing an AI-described API method may find that the method has a different name or different parameters in the actual source code.

Literature and Humanities

An AI summary of a novel might misstate a character’s fate, invent a quote that never appears in the text, or misattribute a line to the wrong author. Because literary analysis often depends on precise quotation, an invented quote can undermine an entire argument.

Business and Economics

An AI report on a company might cite a revenue figure that does not match the actual annual filing, or reference a merger that never closed. It may also describe a market trend using outdated figures presented as current. Business writing that reaches investors or clients carries real financial consequences when a number is wrong.

How Can You Check an AI-Generated Text for Hallucinations?

Check every fact against a primary source, confirm each citation exists and matches its claim, rerun key numbers, and read the text slowly instead of skimming.

The table below breaks this process into 6 concrete steps. Follow them in order for any AI-assisted draft.

Step	What to Check	How to Check	Who Does This
1	Citations	Search the exact title and author in a library database or publisher site	Author/editor (though author must supply new citations in place of fabricated ones)
2	Quotes	Match the quote word for word against the original page	Author/editor (again, the author must find new quotes in place of fabricated ones)
3	Statistics	Trace the number back to the original dataset or analysis results	Author
4	Names and dates	Confirm spelling, titles, and dates against a reliable reference	Author or editor
5	Logic and flow	Check that conclusions actually follow from the evidence given	Author or editor
6	Tone and consistency	Check for style and terminology consistency	Editor

Free tools can support most of these steps. Google Scholar, CrossRef, and publisher websites help confirm whether a citation exists. A DOI lookup tool confirms whether a digital object identifier is real and points to the correct paper. For statistics, going back to the original dataset is the most reliable way to catch a numerical hallucination.

How Can You Prevent AI Hallucinations While Writing a Research Paper?

Prevent hallucinations by breaking your writing process into small, separate AI tasks, verifying each one before moving to the next, and grounding every request in your own sources rather than the AI’s memory.

Most hallucinations creep in when a researcher asks an AI tool to do too much at once. The more a single request depends on the AI’s own recall of facts, the more room there is for invented content. Structuring your workflow around small, checkable steps closes most of that gap before it starts.

Break the Work Into Separate Tasks

• Run literature search, data analysis, drafting, and editing as separate steps, and verify the output of each one before using it as input for the next.
• Avoid chaining tasks in a single prompt, such as asking the AI to find sources and summarize them and draw conclusions all at once, since errors in an early step quietly carry into every step after it.
• Treat each output as a draft to check, not a finished product to paste into your paper.

Ground the AI in Real Sources

• Compile PDFs of the papers you have already selected and chosen to cite, and feed those directly to the AI tool instead of asking it to recall paper content from memory.
• When summarizing or analyzing a source, explicitly instruct the tool to use only the attached document, and spot-check the summary against the original text.
• Avoid asking an AI tool to “find papers on X,” since this relies on its internal recall of citations, which is where fabricated references are most likely to appear.

Avoid Generating Large Sections at Once

• Do not ask a tool to write an entire section or paper in 1 pass; generate a paragraph or subsection at a time so errors are easier to isolate and check.
• Write your own outline and key arguments first, and use AI mainly to help phrase, tighten, or restructure content you already understand, rather than to originate the content itself.
• Review each paragraph immediately after it is generated, rather than waiting until the full draft is finished, since errors are far easier to trace at the point they are introduced.

Know Your Own Data and Field

• Get familiar with your own data and analyses before asking AI for help, so you can immediately spot a number, trend, or result that feels off.
• Keep your own calculations, tables, or summary statistics open in a separate window while reviewing AI-generated text, so you can check claims against them in real time.
• Build enough background in your subfield to recognize a plausible-sounding but incorrect claim, since AI fluency can make a wrong statement read just as confidently as a correct one.

Additional Prevention Tips

• Use retrieval-augmented tools when available (e.g., Paperpal’s ChatPDF feature), since these ground responses in real documents rather than relying purely on the model’s memory.
• Ask the AI to quote directly from a provided source rather than paraphrase, since paraphrasing increases the risk of subtly altering a fact or number.
• Request that the AI flag uncertainty explicitly, and treat any unflagged claim with the same scrutiny, since models do not reliably self-report when they are guessing.
• Keep a running log of every AI-assisted step, including the exact prompt used, so any error can be traced back to where it entered your workflow.
• Cross-check any surprising or convenient-sounding result with a second, independent method or tool before including it in your paper.
• Set aside a fixed verification pass at the end, separate from drafting, dedicated only to checking citations, numbers, and quotes against original sources.

 

How to Spot Data Errors in AI Text

AI-assisted drafts can introduce numbers that drift from your actual dataset, especially when the same figure is repeated across sections. Use the workflow below to confirm every number in the paper before submission.

Step	Action	What to Check	Pass or Fail
1	Locate the source data	Trace every number back to the original dataset, spreadsheet, or output file	Traced or untraced
2	Recalculate key figures	Recompute at least 1 major statistic independently to confirm the method was applied correctly	Matches or mismatched
3	Check the abstract	Confirm every number in the abstract matches the corresponding number in the results, since summary sections are a common place for drift	Matches or mismatched
4	Check the methods section	Confirm sample sizes, variables, and procedures described match what was actually done	Matches or mismatched
5	Check the results section	Confirm every reported statistic, percentage, or test result matches your analysis output	Matches or mismatched
6	Check tables and figures	Confirm that numbers in tables and figures match the numbers stated in the surrounding text	Matches or mismatched
7	Check units and rounding	Confirm units are consistent throughout and rounding does not change the reported conclusion	Consistent or inconsistent
8	Check the discussion	Confirm that any number repeated or reinterpreted in the discussion still matches the original result	Matches or mismatched
9	Confirm internal consistency	Check that the same figure is not reported differently in 2 different sections	Consistent or inconsistent
10	Record the result	Log each figure as verified or corrected before the paper is finalized	Verified or corrected

A number should never be trusted simply because it appears consistently; consistency can mean the same error was copied across sections. Always trace figures back to the original data, not just to an earlier part of the same draft.

What Can a Professional Editor Check?

After the author verifies and revises the paper, it’s a good idea to have a professional editor check grammar, tone, structure, consistency, and formatting, and can flag citations that look incomplete or unusual, but cannot verify facts without the original sources.

A good editor reads with a critical eye for anything that seems statistically unlikely, oddly specific, or inconsistent with the rest of the document. This instinct comes from experience, not from access to the author’s original research, so it works best as a second layer of defense rather than a first one.

• Grammar, punctuation, and sentence structure.
• Overuse of hedging language and modal verbs
• Consistency of terminology, style, and formatting across the document.
• Whether citations follow the required style guide correctly.
• Whether the argument flows logically from paragraph to paragraph.
• Whether citations lead to genuine published research
• Flat or robotic text, or “empty” writing (beautifully phrased, polished sentences that don’t offer any genuine insights)
• Trimming excessive wordiness (AI output often says the same thing again and again in different ways)

An editor without access to the original dataset and analyses cannot verify that the statistics and numerical results are accurate. This means editing is not a substitute for the author’s own fact-checking. Editors can raise red flags, but they cannot close the loop on accuracy alone.

What Must the Author Verify Personally?

Some checks can never be outsourced, no matter how skilled the editor is. The items below require access to the original data and analyses, the original sources, and the author’s own memory of what was actually found during research.

• Locate and open every source cited, rather than trusting the AI’s description of it.
• Confirm that quotes match the original text exactly, including punctuation.
• Check that numbers, percentages, and dates match the original data.
• Confirm that names, titles, and affiliations are spelled correctly and are current.
• Verify that the overall argument does not rely on an invented fact or a misrepresented source.

Only the author knows the intended meaning and the sources actually used during research. This makes personal verification a non-transferable responsibility. Even the best editor cannot know that an AI invented a statistic unless the author flags it or the error is obvious.

Tips for ESL Authors

Writers working in a second language face a specific risk: AI tools often produce fluent, natural-sounding English, which can make hallucinated content feel more trustworthy than it actually is.

• Read the AI output aloud, or use text-to-speech, to slow down reading speed and catch claims that sound too smooth to be true.
• Do not ask an AI tool to generate an entire section or paper at one go, as this increases hallucination risk. Give specific tasks (e.g., “summarize the paper at [URL]”) rather than “write an introduction for a research paper on the relationship between zinc consumption and adipocyte metabolism”.
• Keep a personal glossary of field-specific terms in both languages to check that the AI used the correct technical term, not just a fluent one.
• Choose an editor with expertise in your subject area and ask them to review meaning, not just grammar, since fluency can mask factual gaps.
• Translate key sentences back into your first language to check whether the claim still makes sense and matches what you intended to say.
• Use a second AI tool or search engine to cross-check any fact the first tool provided, since 2 independent sources reduce the chance of a shared error.

None of these steps require advanced technical skill, only a habit of slowing down before submitting a fluent-sounding draft. Building this habit early tends to save far more time later than it costs during the writing process.

Tips for First-Time Authors

Authors publishing for the first time often trust AI output more than experienced authors do, simply because they have not yet seen how confidently these tools can be wrong.

• Treat every AI-generated citation as unverified until you have opened the actual source yourself.
• Manually cross-check your entire reference list before submission (or use a professional editing service that offers reference verification)
• If you’re citing an interaction with generative AI, follow the right format and remember that such citations are treated as weaker as compared to peer-reviewed research.
• Make a table or list of key numerical data for your paper (e.g., sample size) and use it in your prompts.
• Keep a research log that notes which sources you found yourself versus which an AI tool suggested.
• Budget extra time for fact-checking; do not assume a fluent first draft means a finished draft.
• Learn your journal’s or institution’s policy on AI use, since many now require a disclosure statement.

First submissions often set the tone for how supervisors and editors view an author’s future work. Taking the extra time to verify AI-assisted content on a first paper builds a habit that protects every paper that follows.

A Quick Pre-Submission Checklist

Use this table as a final pass before you submit any AI-assisted document to a supervisor, journal, or instructor.

Item	Checked?
All citations located in a real database	Yes / No
All quotes matched word for word	Yes / No
All statistics traced to the original data	Yes / No
All names, dates, and titles confirmed	Yes / No
AI use disclosed per institutional policy	Yes / No
Editor review completed	Yes / No
Final read-through completed without skimming	Yes / No

 

How to Check AI-Generated Code

Verify AI-generated code by running it against known test cases, checking every imported library and function actually exists, and reading the logic line by line instead of trusting that it compiles or runs without errors.

Researchers often treat working code as correct code. A script can run without crashing and still produce wrong results, use a deprecated method, or silently drop data. Code hallucinations are harder to catch than text hallucinations because the errors hide inside logic, not prose. Importantly, code hallucinations can only be caught by authors, not editors, nor even casual collaborators or supervisors.

Common Types of Code Hallucinations

• Nonexistent packages: the AI imports a library or module that was never published, sometimes called “package hallucination,” which can even create a security risk if someone registers that fake package name with malicious code.
• Invented functions or parameters: the AI calls a method that does not exist in the actual library, or uses a real function with parameters that do not match its documentation.
• Silent logic errors: the code runs and produces output, but the underlying calculation, statistical test, or data transformation is wrong.
• Outdated syntax: the AI generates code for an old version of a language or library that behaves differently from the version the researcher is actually using.
• Fabricated benchmarks: the AI reports that a function or model achieves a certain speed or accuracy without this ever being tested.

A Verification Checklist for Researchers

Step	What to Check	How to Check
1	Every import	Confirm the package exists on the official registry (for example, PyPI or CRAN) and is actively maintained
2	Every function call	Compare it against the current official documentation, not the AI’s description of it
3	Output correctness	Test with a small, known dataset where you already know the correct answer
4	Edge cases	Run the code with empty, missing, or extreme values to see if it fails silently
5	Statistical logic	Recalculate 1 result by hand or with a trusted tool to confirm the method is applied correctly
6	Version compatibility	Check that the code matches the language and library versions used in your environment

What a Supervisor/Advisor/Experienced Colleague Can Check

• Whether the code follows standard style and structure for the language.
• Whether variable names and comments match what the code actually does.
• Whether the logic looks reasonable at a high level, based on their own experience.

What the Researcher Must Verify Personally

• That every package and function referenced actually exists and is being used correctly.
• That the output matches an independently calculated or previously published result.
• That randomness is controlled with a fixed seed, so results can be reproduced.
• That the code does exactly what the methods section of the paper claims it does.

• Whether the code is reproducible, meaning it runs the same way on a different machine.

 

Only the researcher who understands the underlying method can confirm that clean code is also correct code.

 

Frequently Asked Questions

How Do You Know If AI-Generated Text Has Fake Citations?

Search the exact title and author in a library database such as PubMed or your university catalog. If it does not appear, or the details do not match, treat it as fabricated until proven otherwise.

Can AI Detection Tools Also Catch Factual Hallucinations?

No. Most AI detection tools identify writing style patterns, not factual accuracy. A hallucinated fact can pass an AI detector completely undetected, so fact-checking must be done separately.

What Percentage of AI-Generated Citations Are Typically Fake?

Rates vary by tool and topic, and no single fixed percentage applies across all studies. Some published tests have found fabricated citation rates ranging from under 10% to over 50%, which is why manual checking remains essential.

Is It Plagiarism If AI Hallucinates a Fact That Turns Out to Be True by Coincidence?

It is not plagiarism, but it is still an integrity risk. Presenting an unverified claim as confirmed research, even if later found accurate, misrepresents your research process to readers.

How Can Students Check AI-Written Text for Hallucinations Without Expensive Software?

Use free tools such as Google Scholar, your library’s database, and publisher websites to verify citations, combined with careful manual reading of every claim and quote.

Do Professional Editors Check Facts and Data?

Editors can catch issues in flow, logic, tone, and accuracy of citations. Fact-checking/data-checking is a separate service that authors should request explicitly if needed.

What Is the Difference Between an AI Hallucination and a Citation Error?

A citation error is a real source with an incorrect date, page number, or formatting. A hallucination is a source, quote, or fact that does not exist at all.

How Should Researchers Disclose AI Use in Their Work?

Most journals and universities now require a brief AI disclosure statement naming the AI tool, the version, and what it was used for, placed in the methods, acknowledgments, or a dedicated disclosure section, since AI tools cannot be listed as authors under current academic policy.

Can an AI Tool Be Listed as a Source in the Reference List?

If the AI output is a stable, shareable link that anyone can access and verify the exact prompts the author used, that AI output can be cited in the paper. APA, MLA, and other style guides have guidelines around this. But note that AI as a source is usually considered much weaker than peer-reviewed research. It’s much better to cite real journal articles, books, or conference papers wherever possible.

Why do AI tools hallucinate instead of saying they do not know?

Because language models generate the most likely continuation of a prompt, and confident text appears far more often in training data than admissions of uncertainty. Most models are not rewarded for abstaining, so a plausible guess becomes the default output whenever reliable information is missing.

Does ChatGPT hallucinate more than other AI models?

No tool is free of hallucination, and rates vary by model version, prompt wording, and subject area rather than by brand. Tools that read real documents before answering generally fabricate fewer citations than tools relying on internal memory, but every output still requires manual verification.

Can you stop ChatGPT from hallucinating by instructing it not to?

No. Instructions such as “only use real sources” or “do not invent anything” give the model no mechanism to check its own output against reality. Reliability improves only when you supply the source documents yourself and restrict the tool to those documents.

What are some real examples of AI hallucination?

Documented cases include legal filings containing invented case law, preprints submitted with fabricated reference lists, and AI generated code importing software packages that were never published. See the examples section above for the full list.

Which parts of a research paper are most at risk of hallucination?

Reference lists, literature review paragraphs summarizing papers the tool was never given, statistics repeated across the abstract and discussion, and any claim about very recent or highly specialized work.

Does retrieval-augmented generation eliminate hallucination?

No. Retrieval confirms that a document exists but does not confirm that the model interpreted it correctly. A grounded tool can cite a genuine paper and still attribute a claim to it that the paper does not make.

Do hallucinations occur in AI models used for data analysis and code?

Yes. Code hallucinations include nonexistent packages, invented function parameters, and silent logic errors that run without crashing. See the section on AI generated code for the full verification checklist.

Is a hallucination the same thing as an outdated fact?
No. An outdated fact was true at some point and can be traced to a real source. A hallucination has no source at any point, which is why it cannot be corrected by updating a citation and has to be removed or replaced entirely.