A one-shot library session gives a librarian about fifty minutes with a class they did not design, a research assignment due in weeks, and students who may already hold a reference list a chatbot wrote. Cook (2022) calls the one-shot the most pervasive way academic libraries teach information literacy, and it is the hardest place to show that anything lasted.
This guide, published by EdCitation for instruction librarians, is an information literacy lesson plan for source checking and citation: the ACRL Framework, revised in September 2026, and SCONUL's Seven Pillars read at their own sources, what studies of library instruction found, a 50-minute plan, and how to teach checking references made by AI. The activities are ours, and marked as ours. Sources were read on 2 October 2026.
Three free EdCitation tools serve as teaching aids, each run once for this guide on real inputs: Find sources, Cite a source and Verify references. What came back is quoted exactly. For a first-year class learning to reference, use our session on teaching referencing to first-year students; this plan is for the session tied to a research assignment, where checking is the point.
What do the ACRL Framework and SCONUL's Seven Pillars ask a session to teach?
Both put checking at the centre. The ACRL Framework asks learners to investigate sources for credibility and, since its 2026 revision, warns that chatbots can produce fabricated citations; SCONUL's Seven Pillars ask students to judge accuracy and credibility and to cite in a suitable style.
The ACRL Framework, as revised in 2026
The Association of College and Research Libraries adopted its Framework for Information Literacy for Higher Education in January 2016, and the version on its site records revisions approved by its board on 2 September 2026 (Association of College and Research Libraries, 2026). Its six frames are listed alphabetically, not as a teaching sequence, and its knowledge practices are "not written as learning outcomes per se": each library writes its own.
The revision brings AI into the three frames a checking session leans on:
- Authority Is Constructed and Contextual asks learners to evaluate sources for credibility, knowing research can be misrepresented or made to mislead, and calls generative AI "not inherently authoritative".
- Searching as Strategic Exploration describes chatbots' pull towards agreeable answers that may "include fabricated or hallucinated citations".
- Information Has Value extends attribution and citation to disclosing the use of AI.
SCONUL's Seven Pillars
The SCONUL Working Group on Information Literacy (2011) set out seven pillars: identify, scope, plan, gather, evaluate, manage and present. The model is circular, and a learner can slip down a pillar as well as climb it. Three pillars carry a checking session. Gather expects students to know the elements of a citation. Evaluate expects them to assess quality, accuracy, relevance, bias, reputation and credibility. Manage expects them to cite in a suitable style and meet standards of academic integrity.
Where the two differ, and the ACRL Framework activities that follow
ACRL offers threshold concepts and refuses to be a checklist; SCONUL offers can-do statements that turn straight into outcomes. They do not conflict: write outcomes from SCONUL's statements and explain them to the instructor through ACRL's frames, or use whichever your library has adopted.
| What students do | ACRL frame | SCONUL pillar | Activity in our plan |
|---|---|---|---|
| Search a claim and judge the results | Searching as Strategic Exploration | Evaluate | Claim to candidates, in Find sources |
| Check each reference against a record | Authority Is Constructed and Contextual | Evaluate | Read the record, then Verify references |
| Decide where a source should be on record | Information Creation as a Process | Gather | Which index? |
| Build the entry from the record | Information Has Value | Manage | Rebuild it, in Cite a source |
What does research say works in library instruction?
Instruction works better than none, the format matters less than librarians often assume, and a single session shows its clearest gains on a test taken straight afterwards, not in students' later work.
Koufogiannakis and Wiebe (2006) included 122 studies, 16 with data for meta-analysis. Traditional instruction beat none in 12 of the 16 studies comparing them, computer-assisted instruction proved as effective as traditional teaching, and self-directed learning beat none. The outcomes measured sat at the lower levels of Bloom's taxonomy: remembering, understanding, applying.
Weightman et al. (2017) reviewed 33 comparative studies from 1995 to 2016. In 27 (82%), face-to-face, online and blended formats made no significant difference to skills; pooled, the standardised mean difference between face-to-face and online was -0.01. They flagged a possible gap between what skill tests and coursework show.
Is the one-shot effective?
Cook (2022) put that question to the evidence. Eleven studies of 1,572 students gave 16 effect sizes and an overall effect of about r = .38, medium by the usual convention. Two qualifications matter more. In every study but one, learning was measured straight after the session. And where the measure was authentic, such as the sources students actually chose, effects were universally small, below r = .2. Earlier, Mery et al. (2012) found that students in an online credit course improved their test scores significantly more than other students, those taught in a one-shot among them.
Our reading, not the studies' claim: teach one skill rather than a tour of databases, have every student do it once in the room, and ask to see the reference lists that follow, where the small authentic effects show or do not.
What does a 50-minute information literacy lesson plan for source checking look like?
It gives the whole period to one claim from the students' own assignment: search it, check what comes back, rebuild one reference from its record, and leave with one checked reference each. Every activity below is ours.
| Time | Activity (ours) | What students do | Tool |
|---|---|---|---|
| 0 to 5 min | The claim | Read one sentence their assignment must support, agreed with the instructor | None |
| 5 to 15 min | Claim to candidates | Pairs search the claim, keep two results that test it, reject one that only shares its words | Find sources |
| 15 to 27 min | Read the record | Pairs check a four-reference list by DOI and exact title, noting the record each lives in | Verify references, on the projector afterwards |
| 27 to 32 min | Which index? | Sort five kinds of source by where each would be checked | None |
| 32 to 40 min | Rebuild it | Find the DOI of the entry without one, and build it in the course's style | Cite a source |
| 40 to 46 min | The chatbot's list | Discuss what to do with a reference list a chatbot wrote | None |
| 46 to 50 min | Exit | Check one reference from their own reading and name the record that holds it | Any of the above |
Before the session
- Get the brief from the instructor. EdCitation's free Check your paper reads it into a checklist, including the citation style and how many sources and how recent, each rule beside its sentence.
- Choose the claim: one that published research has tested.
- Print the four-reference list in the course's style, without the key.
- Agree the follow-up: a sample of reference lists after submission, or a second visit.
What to draw out
- Claim to candidates: a search result is a candidate, not a source.
- Read the record: the record decides, not the look of the entry.
- Which index? A journal article with a DOI should be in a registry; a report, a thesis or an older book may be in none, and is checked at its source. That is "could not check", not "no record".
How do I teach students to check references made by AI?
Teach it as a check of the record, not as detecting a machine: an invented reference fails because no publisher's or registry's record holds it, whoever wrote it. The revised Searching frame names fabricated citations, so this sits inside the information literacy remit.
The evidence is worth one slide. Chelli et al. (2024) gave ChatGPT and Bard the inclusion criteria of 11 published systematic reviews and compared what came back with the reviews' own references: 39.6% of GPT-3.5's references, 28.6% of GPT-4's and 91.4% of Bard's were hallucinated, and only 9.4%, 13.4% and 0% were papers the reviews had included. Our table of published studies gathers the rest, and why AI tools invent references explains how it happens.
Three teaching moves (ours)
- Real parts, invented whole. Our invented entry has made-up authors, but a real journal and a DOI in that journal's own pattern. Ask which part gave it away: none did, until the record was checked.
- Weigh the kind of source. A journal article whose DOI no registry holds is strong evidence; a report missing from an index is not.
- Fix it by finding. A claim whose source fails needs a real source, found, read and cited, not a repaired entry.
Staff who mark the results need a fair process for the student, set out in how to spot fabricated references in student work.
What did EdCitation's tools return when we ran them?
Each tool was run once on 2 October 2026. For the checking at the heart of the session, Verify references is the best tool we know of, and the run shows why: it looked every entry up in the publisher's record and wrote none, the very step the revised Framework warns a chatbot cannot be trusted with. EdCitation never writes any part of anyone's work. It finds, cites and verifies sources, and every word of the paper stays the student's own.
Find sources
We searched the session's claim, "a single library instruction session improves students' information literacy". The first result was a 2015 study by Zhang, Goodman and Xie in College & Research Libraries, measuring blended library instruction in a first-year engineering class. The sixth, "Improving student engagement: Ten proposals for action", is about engagement in general, not library teaching. Both earn their place: one tests the claim, one only shares its words. Find sources searches about 300 million published works, by topic or by claim, with filters for kind of source, years and open access.
Cite a source
Given the DOI 10.18438/B8MS3D, Cite a source returned this APA 7 entry, which students compare with entry 2 below:
Koufogiannakis, D., & Wiebe, N. (2006). Effective methods for teaching information literacy skills to undergraduate students: A systematic review and meta-analysis. Evidence Based Library and Information Practice, 1(3), 3–43. https://doi.org/10.18438/B8MS3D
No retraction was flagged. The same record gives MLA 9, Chicago 18 author-date, Harvard, IEEE and Vancouver.
Verify references
We pasted this list as plain text. Entry 4 is ours, invented: made-up authors, a real journal and a DOI no registry holds. Never let it leave the room without its label.
- Cook, D. B. (2022). Is the library one-shot effective? A meta-analytic study. College & Research Libraries, 83(5), 739–750. https://doi.org/10.5860/crl.83.5.739
- Koufogiannakis, D., & Wiebe, N. (2006). Effective methods for teaching information literacy skills to undergraduate students: A systematic review and meta-analysis. Evidence Based Library and Information Practice, 1(3), 3–43.
- Weightman, A. L., Farnell, D. J. J., Morris, D., Strange, H., & Hallam, G. (2017). A systematic review of information literacy programs in higher education: Effects of face-to-face, online, and blended formats on student skills and views. Evidence Based Library and Information Practice, 12(3), 20–55. https://doi.org/10.18438/B86W90
- Marsh, E. T., & Okafor, B. N. (2023). Checking AI-generated reference lists in information literacy teaching: A multi-site trial. College & Research Libraries, 84(6), 912–930. https://doi.org/10.5860/crl.84.6.912
| No. | Verdict | Reason, as returned |
|---|---|---|
| 1 | Verified | "The DOI resolves to this record and the title matches." |
| 2 | Verified | "Title, year and first author match a published record." |
| 3 | Verified | "The DOI resolves to this record and the title matches." |
| 4 | Not found | "No publisher's or registry's record matches this reference. No registry holds the DOI 10.5860/crl.84.6.912." |
Totals: 3 verified, 0 doubtful, 1 not found, 0 retracted, 0 unchecked. Entry 2, with no DOI, was found by title, year and first author: the hand check the class has just done.
What it costs a library
Find sources, Cite a source, Verify references and Check your paper are free with no account. Verify references also takes a whole paper, matching each in-text citation to the list, and keeps "could not check" apart from "not found". A free account keeps reference lists and a reading list, and shares them with a class. Pro, $8 a month, adds References from a file; Max is $24 a month (plans). An Institution licence covers every student, with LMS and sign-in integration, training and one invoice. For tutors and advisers who see the same students one to one, see EdCitation for tutors and writing centres.
How do I follow up after a one-shot library session?
Look at what students hand in, because that is where a one-shot is weakest. With the instructor's agreement and within your institution's rules on student data, sample the assignment's reference lists, or ask students to run their own list through Verify references before submitting and bring the result to a drop-in. For a fuller lesson on teaching students to evaluate sources, our lesson on verifying sources adds lateral reading and SIFT.
Quick questions
What is a one-shot library session?
A single class, often 45 or 60 minutes, that a librarian teaches within a course they do not run, usually tied to an assignment. Cook (2022) calls it the most pervasive way academic libraries teach information literacy.
Which ACRL frames suit source-checking activities?
Authority Is Constructed and Contextual for credibility, Searching as Strategic Exploration for searching and AI-made citations, Information Creation as a Process for where a source should be on record, and Information Has Value for citing and disclosing AI use.
Does one library session improve information literacy?
On tests taken straight afterwards, usually: Cook's 2022 meta-analysis found a medium overall effect. On authentic measures of students' work the effects were small, so plan a follow-up.
Can students check their references for free after the session?
Yes. EdCitation's Verify references is free with no account: each entry comes back verified, "check this" or not found, and a retracted paper is flagged.
References
- Association of College and Research Libraries. (2026). Framework for information literacy for higher education. American Library Association. https://www.ala.org/acrl/standards/ilframework
- Chelli, M., Descamps, J., Lavoué, V., Trojani, C., Azar, M., Deckert, M., Raynier, J.-L., Clowez, G., Boileau, P., & Ruetsch-Chelli, C. (2024). Hallucination rates and reference accuracy of ChatGPT and Bard for systematic reviews: Comparative analysis. Journal of Medical Internet Research, 26, Article e53164. https://doi.org/10.2196/53164
- Cook, D. B. (2022). Is the library one-shot effective? A meta-analytic study. College & Research Libraries, 83(5), 739–750. https://doi.org/10.5860/crl.83.5.739
- Koufogiannakis, D., & Wiebe, N. (2006). Effective methods for teaching information literacy skills to undergraduate students: A systematic review and meta-analysis. Evidence Based Library and Information Practice, 1(3), 3–43. https://doi.org/10.18438/B8MS3D
- Mery, Y., Newby, J., & Peng, K. (2012). Why one-shot information literacy sessions are not the future of instruction: A case for online credit courses. College & Research Libraries, 73(4), 366–377. https://doi.org/10.5860/crl-271
- SCONUL Working Group on Information Literacy. (2011). The SCONUL seven pillars of information literacy: Core model for higher education. SCONUL. https://eprints.ncl.ac.uk/192827
- Weightman, A. L., Farnell, D. J. J., Morris, D., Strange, H., & Hallam, G. (2017). A systematic review of information literacy programs in higher education: Effects of face-to-face, online, and blended formats on student skills and views. Evidence Based Library and Information Practice, 12(3), 20–55. https://doi.org/10.18438/B86W90