Skip to content

Explicit Instruction: Internet Search

Principle · All principles
Evidence · 13 claims (7 for, 6 mixed) · 31 studies (13 causal, 8 review, 3 quant-synthesis, 3 theoretical, 2 associational, 2 qualitative), q1–q4 · 2 of 31 report an effect size · 6 claims rest on one study

Conditional relationship

This page owns one capability: turning an information need into searches that find relevant material, and revising them when they do not. For an adolescent or adult learner who is a novice searcher on this kind of task, and whose observed starting response is to type the whole question or one vague word, take the first results, and give up or repeat the same query when results are poor, explicit modelling of a short search routine (frame the question, pull out key terms and alternatives, build the query, scan the results, revise for a stated reason), followed by practice on real inquiry questions with feedback on the queries and revisions themselves, is expected to produce more relevant sources found, more deliberate revisions and less abandonment on new search tasks (skill-performance, near-transfer) than leaving learners to explore on their own. The target is a procedure used with judgment; the old page's reason for teaching it explicitly is that unguided exploration does not show learners what to change, which frustrates them.

No claim in the wiki tests this relationship directly. No claim here compares taught search routines with unguided exploration, or varies how query formulation is taught. The nearest evidence is one single-group study inside a claim about fact-checking instruction: eleventh graders given eight lessons on fact-checkers' strategies improved on a task of finding reliable sources in an open internet search, with no control group (Civic Online Reasoning Instruction Improves Evaluation [+M]). The other claims used below are carried from other tasks (writing revision, mathematics and science learning, one-to-one tutoring) and say so where they are used.

The relationship is conditional on the kind of search task (a routine look-up needs little; an open, complex question involves a longer process of forming the question as the learner searches, Task complexity moderates whether users experience the constructive information search process [~M]); on the learner's background knowledge of the topic, which supplies the vocabulary for terms and synonyms; on whether the learner's searches can be seen by anyone who can respond to them; and on the tool (a web engine, a library database, a site search, or a generative AI chatbot that learners may wrongly believe is searching the web).

How this sits beside the converted siblings. Evaluating Sources owns the next judgment, whether to trust what a search returns, including the off-page (lateral reading) move, its dose and its evidence; this page hands over to it at the point of choosing which results to open and does not repeat its model. Inquiry-based Learning holds the investigation in which searching sits. Guided Practice holds the practice phase between a model and independent work, and Scaffolding and Fading how support is withdrawn; the default design below uses both. Direct Instruction holds the general explain–model–guide–hand-over relationship, of which this is one application to a digital skill.

Default design, while the relationship is untested

Each step is a design proposal built from the earlier page's guidance (model turning questions into terms, practise on real inquiry tasks, teach construction and result-checking together, normalise revising a failed search). The evidence status is stated beside each step; none is tested for search instruction as such.

The routine learners carry (a checklist card). Four steps, taught in this order and kept on a card or the first lines of a search log; the same card is the checklist for any assessment of the skill. Untested as a routine; built from the earlier page's guidance.

  • 1. Ask: write what you actually need to know, in one sentence.
  • 2. Pick words: underline two to four key ideas, drop filler words, and list another word for each (a synonym, a narrower or broader term, the word an expert source would use).
  • 3. Search and scan: build the query (add quotation marks for a phrase, OR for alternatives, a minus sign to exclude a meaning), then read titles, snippets and who is behind each result before clicking; skip sponsored results.
  • 4. Check and revise: open the two or three most promising results; if they do not answer the question, say why (too broad, too narrow, wrong words, wrong kind of source), change one thing and search again; keep a source only if you can say why it is trustworthy.

Teaching the steps as the goal, rather than "find a good source", follows Process goals lead to better skill acquisition for novices than outcome goals. [+M], carried from writing revision. Assessment against the card: for a new question, score each step from the learner's log (a stated need; key terms with at least one alternative; a query and a note on the results scanned; at least one revision with a reason, and a reason for each source kept).

Teaching each step. Where a brief names its own instruction sequence (for example short explanation, modelling with think-aloud, guided practice, immediate corrective feedback, independent practice), keep that sequence and run each step of the card through it; the steps below are the default when no sequence is named, and map onto that one (steps 2–4 are explanation and modelling, 5–6 guided practice with corrective feedback, 8–10 the release to independent work).

  1. Watch one search before teaching (5–10 minutes). Give a real question from the learners' current work and ask each learner to find two useful sources for it while logging every query typed, which results they opened, and when they gave up or changed course. This records the starting response. Untested proposal.
  2. Model the routine as a few named steps, thinking aloud (10–15 minutes). On a shared screen, work one question through: (a) say what you actually need to know; (b) underline the two to four key ideas and drop filler words; (c) list synonyms, narrower and broader terms, and the words an expert source would use; (d) build the query; (e) scan titles, snippets and who is behind each result before clicking, noting sponsored results; (f) say what was wrong with the results and change one thing. Teaching the steps of the process as the goal, rather than "find a good source", follows Process goals lead to better skill acquisition for novices than outcome goals. [+M], carried from high-school writing revision and a theoretical chapter, untested for searching.
  3. Teach a few operators only after the terms (5–10 minutes). Quotation marks for an exact phrase, an asterisk for word endings where the tool supports it, OR for synonyms, a minus sign or NOT to exclude a meaning, and a site or domain limit, each shown fixing a problem the class has just seen. Do not teach operators as a list on their own: interfaces change, and question quality and vocabulary matter more (the earlier page). Untested proposal.
  4. Model a failed search and its repair. Run a query that returns poor results, say why (too broad, too narrow, everyday words where the field uses others, the wrong kind of source), change one thing, and compare. Untested proposal; from the earlier page's example of normalising iteration.
  5. Practise on real questions with a search log (20–30 minutes a session). Learners search for questions from their own subject or inquiry work, keeping a short log: question, terms and alternatives, each query, what was wrong with the results, what they changed and why. Asking for the reason behind each revision draws on Prompting learners to self-explain improves understanding and problem solving on immediate tests by a moderate average amount, with little evidence yet on delayed or classroom gains. [+S], carried from explaining worked examples and texts in mathematics and science; untested for searching.
  6. Respond to where the search broke down. Read the log, not only the sources found: a learner whose terms were everyday words needs help finding the field's vocabulary; one whose query was a whole sentence needs to see the key-term step again; one who never revised needs a failed search modelled; one who found good results but opened only sponsored or first-listed links needs the scanning step. Give the least help that makes the next search work. Basis: Contingent scaffolding improves learning more than fixed or absent support. [+M], carried from one-to-one tutoring of mathematics and science; untested for searching and at class scale.
  7. Join searching to judging what it finds. Every practice search ends with a choice of which sources to use and why, so that a fast query that returns misleading sources does not count as success. How to teach the trust judgment is modelled in Evaluating Sources; in two one-session comparisons, learners kept relying on domain names and design after training (One-session checklist (CRAAP) instruction in source evaluation did no better than the alternative methods it was compared with in two quasi-experiments [~M]), so do not treat a URL ending as the test.
  8. Spread it over several sessions and fade the teacher's part. A default arc, untested as a whole: sessions 1–2, the teacher models the full routine on a new question each time and learners practise with a printed planning sheet (question, key ideas, alternatives, query, what changed) on teacher-chosen questions with known good answers; sessions 3–4, pairs search their own inquiry questions with only the revision log, and the teacher reads logs and gives one targeted comment each; sessions 5 and later, individual searches on new questions, at least one in a different tool (a library database, another engine, a site search), without the sheet and with help only on request. Plan four to six sessions of 30–50 minutes across two to four weeks, inside a subject task rather than as a stand-alone skills lesson. The only doses on a claim page here are for fact-checking instruction (two 75-minute lessons at university, six 50-minute lessons in high school, eight lessons in eleventh grade; Civic Online Reasoning Instruction Improves Evaluation [+M]); the counts above are proposals of the same order.
  9. Move a learner on when the process, not only the result, is sound. A learner works without the sheet once they find usable sources for two or three consecutive new questions within the time set, with at least one revision they can give a reason for. A learner whose first observed search (step 1) already shows this skips to step 10 or to the evaluation work. Untested threshold; skipping guidance for those who already succeed follows Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. [~M], carried from cognitive-load studies.
  10. Check later, in a new tool. Two or more weeks after the last session, set a new question in a tool not used in practice and score the log (terms chosen, revisions and reasons) as well as the sources found. Untested proposal.

When a condition is not met: with one session only, teach steps 2 and 4 on one question and give one logged practice search, and say that no lasting change is expected; with no one able to read logs, use worked logs (a model search for the same question, with its revisions and reasons) that learners compare with their own; without open web access, use a library database or a saved set of result pages and teach the same steps on them.

Fitting the design to a situation

The three facts that most change the decision:

  • What the learner does now when searching. A learner who types a whole sentence and takes the first result needs the routine modelled; one who already picks terms and revises needs harder questions and less teaching. If unstated, ask: given a question from their own work, what do they type first, and what do they do when the results are poor?
  • What kind of searching the task needs, and with what tool. A quick fact, an open research question, a professional database and a chatbot each need different moves. If unstated, ask: what will learners search for after this teaching, with which tool, and what will they do with what they find?
  • Who can see a learner's searches, and how soon. Feedback on where a search broke down needs someone, or something, to see the queries. If unstated, ask: during practice, who will look at a learner's queries before their next search, and how?
If the brief says… Then change… Basis
Learners already pick key terms and revise on their first observed search (advanced searchers) Skip the modelling; give harder questions (unfamiliar field vocabulary, specialist databases, conflicting results) and peer comparison of search logs Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. [~M] (cognitive-load studies, carried to searching)
child learners (primary age) Use a child-safe engine or a curated set of sites; teach two steps only (pick the key words, try a different word when it does not work) with the teacher at the screen; no operators Untested; no claim here tests search teaching with children
adolescent learners in a school subject Embed four to six sessions in the subject's own inquiry questions, with the planning sheet early and a revision log later Untested as search teaching; the session counts are of the order used in Civic Online Reasoning Instruction Improves Evaluation [+M] (high-school and eleventh-grade fact-checking lessons, carried)
adult learners returning to study, or with low confidence or literacy Start from searches they already need (health, benefits, housing, jobs, their children's school); model typing and reading results step by step; let them say terms aloud before typing; keep result pages short and reading load low; praise a revision, not only a find Untested; from the earlier page's target learners
adult learners in a work setting (workplace-clinical, office, trades) Practise on the tools and questions of their job (a clinical database, an internal knowledge base, a supplier catalogue); use the field's own vocabulary lists as the source of terms; check with a work question weeks later Untested; carried from the earlier page
A discipline's literature search (health sciences, law, pharmacy) Teach the field's database and its subject headings; consider a game or competition format for practice, but score the quality of the search strategy, not only completion Game-based learning improved pharmacy students' literature searching course scores in one reported study [+W] (second-hand report; credits were given for task completion, not search quality)
The task is a quick look-up, not an open question Teach the key-term and scanning steps only; do not require a full log Task complexity moderates whether users experience the constructive information search process [~M] (workers' information seeking; routine tasks were straightforward, carried to learners)
The task is an open, complex research question Expect the question to change as learners search; build in a point where they rewrite the question from what they have found, and log both versions Task complexity moderates whether users experience the constructive information search process [~M] (complex tasks involved construction and formulation over time, carried)
online-self-paced, no teacher present Short recorded demonstrations of each step and of a failed search being repaired; practice questions where the learner submits their queries and reasons before a worked model log for the same question is shown; a delayed check in a new tool Untested; carried from the default design
Mixed mode: some learners remote, some in the room, or phone-only Run the modelled search on a shared screen visible to both groups; have remote and phone users paste their queries into the shared chat or a shared document so the teacher sees all of them; on phones, practise with voice search as well as typing and keep the log to two lines Untested proposal
One teacher, a large class Have every learner post their first query for the same question at once (a shared document or poll); discuss the commonest problem with the whole class; read only the logs of learners who found nothing usable Untested; contingent-support evidence is one-to-one (Contingent scaffolding improves learning more than fixed or absent support. [~M])
Scarce facilitator time (volunteers, one tutor checking in briefly) Give a one-page worked log for each practice question; the facilitator checks one thing per learner: whether the last revision had a reason Untested proposal
Two sessions only (for example two 45-minute lessons) Session 1, card steps 1–3 (question, words, query, reading results): 5 minutes explanation, 10 minutes modelling with think-aloud on one question including a failed search, 20 minutes guided practice on a set question with the card and immediate correction of each learner's first query, 10 minutes independent search on a second question. Session 2, card step 4 and judging what was found: 5 minutes recap from session-1 logs, 10 minutes modelling checking and revising (including leaving a page to see who is behind it), 20 minutes guided practice on the learners' own questions with corrective feedback on revisions and source choices, 10 minutes assessed independent search on a new question, scored with the card. Do not expect retention beyond the sessions Untested split; the evaluation part draws on Evaluating Sources, and one-session information-skills teaching gave only modest gains (One-session checklist (CRAAP) instruction in source evaluation did no better than the alternative methods it was compared with in two quasi-experiments [~M], carried from source evaluation)
The brief names its own instruction sequence (explanation, modelling, think-aloud, guided practice, corrective feedback) Keep the sequence and its order; teach one or two card steps per pass through it rather than replacing it with this page's steps; give the corrective feedback on the query and the revision, not only the source found Untested; feedback that responds to where the search broke down is carried from Contingent scaffolding improves learning more than fixed or absent support. [+M] (one-to-one tutoring)
A single session only (a library visit, an orientation) Model one search including a failed one and its repair, then one logged practice search on the learners' own assignment question; give the course teacher the planning sheet to reuse Untested; the one-session evidence in One-session checklist (CRAAP) instruction in source evaluation did no better than the alternative methods it was compared with in two quasi-experiments [~M] is for source evaluation, where one session of either method gave modest gains
A fixed short time budget (for example 20 minutes) 5 minutes modelling the key-term step, 2 minutes on one failed search and its repair, 10 minutes of learner searching on one question with a two-line log, 3 minutes comparing one log with a model Untested proposal
The brief names its own search routine (a library's search worksheet, PICO, a database's own guide, a company template) Keep it and its categories; add only what the evidence here says matters: modelled examples, a stated reason for each revision, and feedback on the queries Untested; the reason-giving step is carried from Prompting learners to self-explain improves understanding and problem solving on immediate tests by a moderate average amount, with little evidence yet on delayed or classroom gains. [+S]
Learners search on private or sensitive topics (health, immigration, money, identity) Let the learner choose whether anyone sees their log; practise the routine on a neutral question in class and let them apply it privately; teach clearing history on shared devices Untested proposal
Learners use a generative AI chatbot in place of search Show what the chatbot did and did not do (it may not have searched anything); have learners run the same question as a search, compare, and trace any claim to a source they found Learners' belief that chatbots search the web: Undergraduate students' genAI knowledge contains specific confident misconceptions, most prevalently that AI chatbots perform web searches [~M] (one undergraduate sample; describes beliefs, tests no teaching). The comparison activity is untested
The search tool shows AI summaries above results Ask learners to open and check at least one source behind the summary before using it; log whether the summary changed what they searched next AI overviews in an academic search engine showed mixed effects: summaries may reduce mental demand and frustration for some users, with slightly fewer clicks and query reformulations [~M] (one within-subjects user study; mixed effects, fewer reformulations)
The goal is a whole inquiry product for a real audience (a report, a brief, a presentation) Keep the product whole; teach searching inside it at the point the learners need sources, and assess the search log alongside the product rather than as separate items Untested proposal
Instruction or sources are in the learners' second language Let learners list terms in their first language first, then find the second-language and field terms; show that changing the search language changes results Untested proposal
No open web access, locked devices, or an offline exam Teach on a library database or saved result pages; assess with a planning sheet and a set of printed result lists from which learners choose and justify a revision Untested; carried from the earlier page's point that durable principles matter more than one interface
High stakes, or the skill is graded Grade on a new question in an unfamiliar tool, scoring the terms, revisions and reasons as well as the sources found Untested proposal

Twelve of the twenty-four rows cite a claim; none of those claims tests search instruction. The game-based row rests on a second-hand report about literature searching, the two task-type rows on a theoretical account of workers' searching, and the two AI rows on one belief survey and one user study of search interfaces; the expertise, adolescent, class-size, two-session, single-session, named-sequence and named-routine rows carry findings from other tasks (cognitive-load studies, fact-checking lessons, tutoring, one-session source evaluation, self-explanation), as each says, and their design moves are untested. The other twelve rows are untested proposals.

Observation, state and explanation

What can be observed is a search under stated conditions: the question given, each query typed, the time between queries, the results opened, the reason given for each change, the sources chosen, and when the learner stopped. "Is information literate", "has search skills" or "is frustrated by searching" are inferred states. A good source found does not show a sound process: it can come from luck, a helpful autocomplete or a single well-known site.

Response Plausible explanations Observation that could change the interpretation
Types the whole question as a sentence and takes the first results Knows no other way; the engine usually answers sentences well enough; did not see a reason to do more Give a question where the sentence query returns poor results, then ask for key words; if they pick good terms at once, the routine was absent, not the skill.
Repeats the same query, or near-identical ones, then gives up Lacks alternative vocabulary for the topic; does not know which part of the query to change; frustration or low confidence Ask for three other words for each key idea; if they cannot, the issue is topic vocabulary, so supply a source of terms; if they can, model changing one part.
Uses operators correctly but finds little that is relevant Operators learned as tricks; the underlying question is unclear; terms are everyday rather than the field's Ask the learner to say in one sentence what they need to know, then compare their terms with those in a known good source.
Finds relevant pages fast but uses an unreliable or sponsored one Search worked, judgment did not; takes ranking as a sign of quality Give two results on the same topic with unequal backing and ask which to use and why; route to the evaluation work if the reasons rest on rank or design.
Searches well in class, poorly in a new tool or on their own assignment The routine is tied to the practice tool or the planning sheet; low stakes made it seem not worth the time Set a matched question in a new tool with and without a one-line prompt to use the routine; compare.

These rows are untested proposals. Each probe is itself practice in searching, so record what was asked and keep the explanations that remain open.

Evidence and qualifications

No claim in the wiki compares explicit search instruction with unguided exploration, or tests any step of the default design for searching. The claims below test neighbouring relationships and are carried to this one.

  • Civic Online Reasoning Instruction Improves Evaluation [+M]: three studies of teaching fact-checkers' strategies (a high-school field study with matched controls, a randomised pilot in four university sections, and a single-group study of 68 eleventh graders), none with an effect size. The one result that touches searching is in the last study: after eight lessons, scores rose on three of four tasks, one of which was finding reliable sources in an open internet search; with no comparison group the gain cannot be separated from practice or maturation. The claim is about judging credibility, so it bears on step 7 and on the doses in step 8, not on query formulation. Its recorded evidence (three studies, one at q3) would allow a strong marker; it is cited here at moderate because only one uncontrolled study touches searching.
  • Process goals lead to better skill acquisition for novices than outcome goals. [+M]: high-school students learning writing revision did better with goals about the strategy than with goals about the final score, and a theoretical chapter synthesises studies (including dart-throwing) to the same effect; the experiment also suggested shifting to outcome goals once the skill is mastered. No effect size printed. It is the basis, carried, for making the search steps the goal early and the result the goal later; it is not a study of searching.
  • Prompting learners to self-explain improves understanding and problem solving on immediate tests by a moderate average amount, with little evidence yet on delayed or classroom gains. [+S]: a meta-analysis of 69 effect sizes from 64 reports found a moderate average benefit of self-explanation prompts (g = .55) across task types, subjects and levels; a mathematics meta-analysis finds evidence after a delay or in classrooms much more limited. It is the basis, carried, for asking a reason for each revision; no study here tests reason-giving during searching, and a learner without the topic vocabulary may explain a poor revision confidently.
  • Contingent scaffolding improves learning more than fixed or absent support. [+M]: fully contingent one-to-one tutoring of fourth and fifth graders in long division beat fixed, partly contingent and no support at immediate and one-month tests (8 per condition); a meta-analysis ranked explicit strategy training above scaffolding and scaffolding above coaching; a review notes most evidence is one-to-one. Its studies table gathers 20 distinct studies across the claim and 17 narrower ones. It supports responding to where a learner's search broke down, carried from mathematics and science; its ranking of explicit strategy training first is consistent with modelling the routine before coaching it.
  • Task complexity moderates whether users experience the constructive information search process [~M]: a theoretical paper reports that workers sought information straightforwardly in routine tasks but went through a process of construction and formulation in tasks they saw as complex, which took longer. It qualifies the relationship: a full routine with a log suits open questions, and for routine look-ups it may be overhead. Workers, not learners; no instruction tested.
  • One-session checklist (CRAAP) instruction in source evaluation did no better than the alternative methods it was compared with in two quasi-experiments [~M]: limits the model. In one-shot library sessions with first-year business and nursing students, a checklist and an alternative method did not differ on quizzes or accuracy, both gave modest gains, and learners in both groups kept leaning on domain names and design. It concerns evaluation rather than searching, but it is the only evidence here on one-session information-skills teaching, and it suggests that a single session of explicit teaching changes habits only partly, whatever is taught. It also warns that search results judged by surface cues stay misleading however good the query.

Learners, tasks and settings differ across these claims (writing revision, mathematics tutoring, workers' information seeking, credibility judgments), so do not rank them by effect label. None measures search performance, and none follows learners beyond a month for searching.

The evidence by question

Further evidence, not yet read against this model

Claims the earlier page cited that are not used in the model above, and neighbouring claims found while rewriting it. None has been re-read against the model.

Objective and learner-valued goal

The designer's objective is usually that, given a new question in an unfamiliar tool, the learner can find relevant, usable sources in reasonable time and say why each revision was made, weeks after teaching. Elicit separately what the learner wants from searching: an answer to a question that matters to them now, an assignment finished with acceptable sources, less time wasted, or not feeling lost in front of others. Agreement and divergence are both observations. A learner who values a fast answer may stop at the first plausible result however well they can search; that is a choice to discuss, not a missing skill.

For example, an adult in a returning-to-study course may want to find out whether they qualify for a childcare grant, while the course objective is locating academic sources for an essay. Using the grant question as the first modelled search serves the learner's goal and teaches the routine; the essay search in a library database is then the check of transfer to the course objective. Record both, so that success on the grant search is not read as the capability the essay needs.

What would revise this model?

The expectation should weaken if comparisons of taught search routines with unguided exploration, on matched learners and new search tasks, show no difference in sources found, revisions or abandonment, or a gain that disappears within weeks or in a new tool. If teaching operators alone matches the full routine, the routine is overhead. If asking for reasons slows learners without improving their searches, drop the log for all but the hardest tasks. If generative tools that answer directly make query revision rare in learners' own work, the target should move from building queries to checking what an answer rests on. Do not protect the model by calling every failure too little practice or too little background knowledge after the fact.

Finding relevant sources on a practice question, choosing trustworthy ones, searching well in a new tool, and searching well weeks later in the learner's own work are separate claims. The evidence here reaches none of them directly for searching; it establishes no number of sessions, no routine and no hand-over criterion.

Claims

Examples

  • Query expansion exercises, keyword extraction from a problem statement, and guided result comparison
  • Web Search Scavenger Hunt — constrained practice on query generation and retrieval
  • Boolean Operators — a narrower strategy inside the broader principle
  • Instructors modeling a failed search and then revising it can normalize iteration and reduce frustration
  • Direct Instruction: Keyword Search — selecting, combining and refining keywords, with credibility checks on what is retrieved
  • Search Strategy Scaffolds — structured supports for formulating queries and evaluating results
  • Lateral Reading — the off-page check that follows a search, owned by Evaluating Sources

Key Sources

  • Afflerbach, P. A., & Cho, B. Y. (2010). Determining and describing reading strategies: Internet and traditional forms of reading. In Metacognition, strategy use, and instruction.
  • Gross, M., & Latham, D. (2013). Addressing below proficient information literacy skills. Library & Information Science Research, 35(3), 181-190. doi:10.1016/j.lisr.2013.03.001
  • Kingsley, T., & Tancock, S. (2014). Internet inquiry. The Reading Teacher, 67(5), 389-399. doi:10.1002/trtr.1223
  • Roberts, L. (2017). Research in the real world. College & Research Libraries, 78(4), 527. doi:10.5860/crl.78.4.527
Page metadata
Field Value
type principle
id explicit-instruction-internet-search
title Explicit Instruction: Internet Search
description For a novice searcher who types whole questions and stops at the first results, explicit modelling of a routine for turning a question into terms, building and revising queries and checking results, practised on real inquiry questions with feedback on the queries, is expected to find more relevant sources with less abandonment than unguided exploration, though no claim in the wiki tests search instruction directly.
status review
canonical True
generated by claude/unspecified on 2026-10-14
trust unverified — no one has confirmed this page's content
sources 1 — gross-2013

Frontmatter is the page's machine-readable record — see the Schema & Guide.