Industries / Government, education and community / Government / Libraries
Programs and teaching
What libraries do beyond lending: story times and reading programs for children, teaching students to find and judge information, classes and maker spaces, help with jobs and social services, and the building as a place to gather.
7
Milestones achieved, from 1896 to 2025.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1896
A children's room by design
The Pratt Institute Free Library in Brooklyn opened in May 1896 with a children's room built into its plan, where children borrowed books and attended story hours; Pratt calls it "the first to be part of a library's design". In its first year 1,753 patrons under 14 registered and children borrowed 39,007 of the 255,277 books lent. Anne Carroll Moore ran the room for ten years, then moved to the New York Public Library and planned the children's collection for its main building, opened in 1911.
LessonThe new work needed its own training, and Pratt added a course in children's librarianship in 1899.
Credit
Mary Wright Plummer, library director, and Anne Carroll Moore, Pratt Institute
1993
School librarians linked to scores
A Colorado study of 221 of the state's 1,331 public schools found that the size of a school library's staff and collection was "the best school predictor" of test scores, after allowing for community poverty and other school factors. The Colorado State Library published it in 1993, and its research unit says "more than a dozen" later state studies found similar links.
LessonThe design shows a link, not a cause, and the evidence did not protect the jobs: full-time-equivalent US school librarian posts fell from more than 52,500 in 2009–10 to just over 39,300 in 2022–23, a 25% drop, by Lance, Debra Kachel and Caitlin Gerrity's count from federal data.
Credit
Keith Curry Lance, with Lynda Welborn and Christine Hamilton-Pennell, Library Research Service, Colorado State Library
2009
Library hires a social worker
San Francisco Public Library hired Leah Esguerra, a marriage and family therapist, in 2009 to connect patrons inside the library with housing, food, health and benefits services. PBS NewsHour called her "the nation's first full-time library social worker", and reported in January 2015 that about 150 formerly homeless patrons had found permanent housing through the program.
LessonOne post grew into a team: by the library's own count, repeat outreach contacts rose from 8,247 in fiscal year 2023 to 14,671 in fiscal year 2026.
Credit
San Francisco Public Library; Leah Esguerra
2010
Summer reading study lacks controls
A three-year study at 11 US sites, funded by the Institute of Museum and Library Services (IMLS), reported that children in public library summer reading programs scored higher on reading tests than children who did not join. Its own tables tell a weaker story: the joiners already scored higher before the summer (663 against 601), the autumn difference was not statistically significant, and only 149 of more than 800 invited pupils could be matched at the end.
LessonFamilies chose whether to take part and, as an independent review noted, "no control group was established", so the study cannot show that programs which 95.2% of US public libraries run cause the gains.
Credit
Susan Roman, Deborah T. Carran and Carole D. Fiore, Dominican University
2016
Framework replaces competency standards
The Association of College and Research Libraries (ACRL), the US association for university and college librarians, adopted its Framework for Information Literacy for Higher Education on 11 January 2016 and rescinded its 2000 competency standards on 25 June 2016. Information literacy is the skill of finding, judging and using information, and the Framework teaches it through six broad ideas, called frames, such as "Authority Is Constructed and Contextual" and "Searching as Strategic Exploration", in place of a checklist of skills.
Credit
ACRL Board of Directors
2018
A library built for doing
Helsinki's Central Library Oodi opened in December 2018 and on 27 August 2019 won the IFLA/Systematic Public Library of the Year award from 16 applicants. IFLA is the International Federation of Library Associations and Institutions. By Oodi's own figures it had over 2.7 million visits in 2025, about 1,200 free events, 19,000 uses of equipment in its Urban Workshop and about 550,000 loans.
LessonWith about five visits for every loan, the building is measured by how people use its space and events, not only by borrowing.
Credit
Helsinki Central Library Oodi, City of Helsinki
2025
Courts reverse IMLS shutdown
IMLS, the main federal funder of US libraries, was told by Executive Order 14238 of 14 March 2025 to cut itself to the minimum the law requires, and on 31 March its staff of about 70 were put on leave. A federal court in Rhode Island, in a suit by 21 states, blocked the cuts on 13 May 2025 and made the block permanent on 21 November 2025. The administration withdrew its appeal on 6 April 2026, and on 9 April settled a separate suit by ALA and AFSCME, a public employees' union, reinstating terminated grants.
LessonA court can keep an agency alive but cannot fund it: for fiscal 2027 the White House asked for $6 million "to shut down the agency", and a House subcommittee approved $291.8 million in June 2026.
govinfo.govwfae.orgpublishersweekly.compublishersweekly.comala.orgala.orgala.org
Credit
21 state attorneys general; the American Library Association; AFSCME; Judge John J. McConnell Jr
Watch for
Microsoft and PLA AI training
Microsoft and the Public Library Association (PLA), ALA's public library division, announced "Building AI Skills and Agency Through Libraries", offering free AI training, webinars and a community of practice to library workers and patrons across the US. The announcement gives no number of libraries, no funding amount and no outcome measures.
Achieved whena count of libraries and people trained is published, with an outside evaluation of what participants can do afterwards.
Claimed by
Microsoft and PLA, October 2026
Singapore's AI reading program
Singapore's National Library Board and KPMG launched "Read to Lead: Building an AI-Ready Mind", a year of talks, library activities and a toolkit meant to reach more than 2,000 working professionals. Their own poll of 1,150 professionals found only four in 10 confident at telling accurate content from AI-generated misinformation.
Achieved whenreach is reported and someone outside the two partners measures a change in participants' ability to spot misinformation.
Claimed by
National Library Board and KPMG, July 2026
Revised Framework adds generative AI
ACRL's Framework page says its board approved revisions on 2 September 2026, and the text now says learners should "understand that emerging technologies, such as Generative AI, are not inherently authoritative".
Achieved whenuniversity libraries are shown to use the revised text in routine teaching, for example in an independent survey or in published course materials.
Claimed by
ACRL, September 2026
Open
01
Count library programs by topic
A public, machine-readable dataset of what US public libraries actually run: 12 months of events from the online calendars of 100 library systems, many of which use LibCal, Communico or BiblioCommons Events, each labelled by an AI model with a topic such as story time, job help, benefits help, digital skills, English classes, maker space or teen space. Publish the data and the code.
The briefCredit · unclaimed02
Rate AI-made literacy exercises
120 short information-literacy exercises, 20 for each of the six frames of the revised ACRL Framework, drafted by an AI model. Three instruction librarians rate each one blind for fit to its frame and for accuracy, and rate 30 existing librarian-made exercises the same way for comparison. Publish the set and the ratings under an open licence.
The briefCredit · unclaimed
Open 01 · the brief
Count library programs by topic
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, machine-readable dataset of what US public libraries actually run: 12 months of events from the online calendars of 100 library systems, many of which use LibCal, Communico or BiblioCommons Events, each labelled by an AI model with a topic such as story time, job help, benefits help, digital skills, English classes, maker space or teen space. Publish the data and the code.
Why it is still open
The federal Public Libraries Survey for fiscal 2024, answered by 8,931 of 9,244 public libraries, counts program sessions only by audience age and by format (onsite, offsite, virtual), with no topic. ALA's national study of library programs (NILPPA) gives only IMLS's figure of 5.2 million programs in 2016, and PLA retired its Project Outcome platform in August 2025. Delaware's state library processes its libraries' LibCal exports into yearly counts, and Chicago and Austin publish feeds of upcoming events, but this research found no dataset of program topics across many systems.
Built with what exists
a Browser Use agent to collect calendar pages (see the Note "Deleting 80% to start over"); Claude, Gemini or GPT to label events, or Gemini Decision 1 to choose one topic from a fixed list for each event (see the Note "A thousand questions, one pass"); the public Public Libraries Survey files to check totals.
First step
Choose 10 systems of different sizes from the fiscal 2024 survey file, record which calendar software each uses, and hand-code 50 events from each.
Done looks like
events for 100 systems published; topic labels match two human coders on at least 90% of a random 1,000-event sample; scraped session counts within 10% of hand counts for 10 systems; each system's total compared with its survey total and the gap reported.
Open 02 · the brief
Rate AI-made literacy exercises
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
120 short information-literacy exercises, 20 for each of the six frames of the revised ACRL Framework, drafted by an AI model. Three instruction librarians rate each one blind for fit to its frame and for accuracy, and rate 30 existing librarian-made exercises the same way for comparison. Publish the set and the ratings under an open licence.
Why it is still open
ACRL's Framework Sandbox, where members share teaching materials, describes no review or scoring of materials against the frames, and the revised Framework text dates only from 2 September 2026, so existing sets were written for the older text. A search for studies in which librarians rated AI-drafted information-literacy exercises found case studies of librarians using ChatGPT's output as material for students to check, not as exercises to be rated.
Built with what exists
Claude, Gemini or GPT to draft; Gemini Decision 1 to run a fixed checklist of rubric questions on each exercise and compare its results with the librarians' ratings (see the Note "A thousand questions, one pass").
First step
Draft two exercises per frame, 12 in all, from the revised text, and recruit three raters.
Done looks like
120 exercises published; agreement between raters reported as Krippendorff's alpha, a standard agreement score; the share rated "aligned" by at least two of three librarians, for the AI-made and librarian-made sets side by side.