Attention literacy: what deserves our attention?

Artificial intelligence (AI) is supposed to help us deal with information overload.

It has certainly helped us get quick answers, but it doesn't mean that it is the right answer, because of ease of generation, we are witnessing another spike in volume and veracity of algorithms to promote. Are we thinking about information overload in the right way?

Are we reaching peak AI?

When participants at a recent knowledge cafe were asked if they felt it challenging to keep up to date and overwhelmed by the amount of information, it was a resounding 'yes!'.

There is simply too much to follow: new tools, research papers, webinars, free courses, vendor announcements, newsletters, social media posts and almost daily claims about the next revolutionary AI feature.  The question then focused on how to keep up to date with what matters.

Participants described becoming increasingly selective. Long AI-written LinkedIn posts were being skipped. Emails offering yet another free AI course went unread. AI-generated adverts were ignored.

This is not necessarily disengagement. It may be a sensible response to information abundance. Oh and Park's (2026) study, The Impact of Generative AI-Based Information Overload and AI Literacy on Individual Innovation Capability, offers an interesting parallel. Greater AI-related information overload was associated with more information avoidance, while avoidance was associated with stronger creative problem-solving. The authors suggest that deliberately filtering out unnecessary information can help people focus on what matters. That is different from simply switching off, and perhaps that is where Knowledge and Library Services have something important to contribute.

Our value has never been simply about finding more information. It is about helping people find information that matters to them.

AI is creating the problem it is being asked to solve

We know that there is a circular relationship between AI and workload.

Generative AI makes it much easier to produce a report, email, application, complaint, or social media post. That can be empowering, but if everyone can create more content, everyone else has more content to process.

AI is then offered as the solution: summarise it, triage it, extract the key points. So, we potentially end up using AI to manage an information problem partly created by AI. Should we then be teaching people when to stop producing information?

Knowing what to ignore

Information literacy helps us find, evaluate, and use information. Attention literacy adds the screening question "Does this information deserve my attention?"

That might include knowing:

• What can safely be ignored
• When we have enough evidence
• When to stop searching or prompting
• When a summary is sufficient
• When we need to read the full source
• When shortening something has removed essential nuance
• How algorithms and personalisation influence what is put in front of us

This is not a licence to ignore inconvenient evidence: the challenge is to filter deliberately without losing what we need to make a sound judgement. AI literacy goes beyond knowing how to use a tool. It needs to include recognising and regulating our own thinking. Only by doing this can we use AI safe and effectively to support a task.

There is an inevitable tension between efficiency and thinking. If we stop reading full content, does this impact on our ability to think and learn. The National Literacy Trust's AI and Literacy provides useful wider context around AI, literacy and critical engagement.

If productivity is measured primarily through visible outputs, reading, reflecting and thinking can look unproductive. AI's ability to provide an instant summary potentially reinforces that.

Sometimes a summary is exactly what we need.  We also need to know when to read things in full.

The old skills still work

For all the discussion about new AI literacy skills, we repeatedly came back to very familiar territory.

Before relying on an AI-generated answer, we can still ask:

• Can I verify it?
• Does it point me to an identifiable source?
• Is the source authoritative and current?
• Can I trace the claim back to the original research or guidance?
• Does the information actually help answer my question?

We can check whether references exist. We can open the paper. We can verify bibliographic details in established databases. We can look for evidence that challenges the answer we are being given. And we can ask what might be missing.

We can broaden how we apply this. A BBC Verify example shared in the Café involved images apparently showing the Pink Panther taking part in riots in France. They were AI-generated. At the less consequential end of the spectrum, Just Eat apologised after AI-enhanced menu images included some spectacularly odd interpretations of food.

The House of Commons Library briefing on AI content labelling is a helpful introduction to approaches including labels, metadata and watermarking.

We need to become critical friends to the AI model by instructing what not to do. Also, the people using it to be more critical of their use and response. Be more accountable for their actions.

So where does this leave library staff?

AI does not diminish the importance of Knowledge and Library Services. If anything, it strengthens the case for what we already do.

We can help people:

• Distinguish finding information from evaluating it
• Trace claims back to original sources
• Understand where AI-supported retrieval may be incomplete
• Recognise fabricated or inaccurate references
• Deliberately seek counter-evidence
• Evaluate new AI research tools
• Build reproducible testing approaches
• Develop selective rather than overwhelming current awareness
• Understand when a summary is enough and when they need the original
• Protect reading and thinking as legitimate parts of professional work

We are seeing more services issuing statement on AI use, alongside training offers, for example, Somerset NHS Foundation Trust Libraries’ guide to Generative Artificial Intelligence in Evidence-Based Practice, Use of AI in evidence generation: NICE position statement and NHS Scotland's AI Literacy.

It is worth taking a few minutes to reflect on how our own information habits are changing and identify 1 small improvement we could make for our customers.

Taking Oh and Park's emphasis on self-reflection as a starting point, we might build a brief pause into AI training: What do I need to know? What can I set aside? What still needs checking? Perhaps our contribution is not another stream of information, but helping people make room for the information, and the thinking, that matters.