Research notes

TEF Canada writing (EE) and speaking (EO): prompt bank, what repeats, and templates

Checked 4 Nov 2026. Scope: the two writing tasks (EE Section A and B) and the two speaking tasks (EO Section A and B) of TEF Canada in its current format. Every prompt below is paraphrased in one line. Confidence labels: high = official LFA document, read in full for this report; medium = a prep site's bank that says or implies its prompts come from real sittings, or several secondary sources that agree; low = one secondary source, unsourced claims, or my own inference.

0. Key points for the study plan

1. The tasks are fixed and fully predictable (high). Writing takes 60 min: Section A is 25 min, at least 80 words, in several paragraphs, continuing a news-in-brief story. Section B is 35 min, at least 200 words: a letter to a newspaper reacting to a quoted statement, with at least 3 arguments. Speaking takes 15 min: Section A is 5 min, a formal phone call about an everyday ad in which you ask about ten questions. Section B is 10 min, an informal chat in which you present an ad to a friend and try to persuade them. The friend (the examiner) is trained never to agree. Sources: LFA EE sample, LFA EO sample, 2024 examiner handbook, exam-day page.
2. I collected 301 distinct prompts: 54 for EE-A, 49 for EE-B, 108 for EO-A and 90 for EO-B. The target was 40+ per section.
   • Real candidate-reported prompts were reachable only through prep-site banks, mainly OPAL (tefcanada.opal-ca.com, "sujets d'actualité"), French Exam Hub (prompts "paraphrased from candidate discussions", 2025–26) and Mocko ("recurring in the question pool").
   • These banks copy one another and mix in the official samples. Treat them as the circulating pool, not as verified live papers (medium-low; see §1).
3. What repeats most (counts in §4):
   • EE-A: animal surprises are the biggest family (12 of 54 prompts, 22%). Crime or dispute stories repeat most often. The official "80-year-old vs car thieves" story was seen 5 times, and "sausage-seller row ends in prison" 4 times.
   • EE-B: absolute statements about technology and screens (9) and about society and values (9) lead. "Package tours vs real discovery" (6 sightings), "social networks are a permanent danger for children" (5) and "competition has made people selfish" (5) top the list.
   • EO-A: five ad families cover 75% of prompts (81 of 108): thrill and outdoor activities, home and event services, jobs and castings, outings and events, and courses. Casting calls for extras appear as 6 different Section A ads.
   • EO-B: sport, leisure and thrill activities (20), travel (13), buying or using a product (12), volunteering (11), courses (10) and environmental or community action (10) cover 84%. About half the Section B prompts give the friend a profile you can build arguments on, such as "has free time", "is looking for work" or "loves animals".
4. All 16 official LFA sample prompts (4 per section) reappear inside OPAL's "sujets d'actualité" bank. The 2016 manual already printed all 8 writing samples and 4 of the 8 speaking samples word for word. Drill the official samples first: they are guaranteed to be on-format, and the main candidate bank treats them as live subjects.
5. Templates give the best return because the task shapes never change. One question checklist covers 108 of 108 EO-A prompts. One 1+3+1+1 letter frame covers 49 of 49 EE-B prompts. One present–persuade–handle-objections loop covers 90 of 90 EO-B prompts. One 3-paragraph news frame plus 8 small vocabulary packs covers 54 of 54 EE-A prompts (§5).
6. NCLC 7 needs a solid B2. Writing NCLC 7 is 428–471 out of 699, about mid-B2. Speaking NCLC 7 is 456–493, the upper half of B2 (LFA NCLC chart). On the official speaking grid, B2 means suitable questions with confident reactions to unexpected answers (Section A), and a clear, detailed presentation with arguments that are illustrated and firmly defended (Section B). See the grid summary in tef-format-scoring.md §4.2, from the 2025 grid.

1. Sources and how far to trust them

Limits of this run. The shared web-search budget ran out before this task's first search. Everything was found by fetching known pages, reading site sitemaps and following links. Sources I could not reach:

• Reddit: blocked for both the fetch tool and the browser pane.
• CanadaVisa forum: behind an Imperva bot check, which I did not bypass.
• Facebook groups: not reachable.
• Telegram: guessed channel names led only to contact pages. One small channel, @tef_canada, only reposts the official samples.
• YouTube: pages return no video titles to the fetch tool.
• tefcanadaonline.com: its subject bank is paywalled. The site advertises 74 EO subjects (34 A + 40 B) and 61 EE subjects (19 A + 42 B) (EE page).

No first-hand candidate testimony was reached. A follow-up run with search budget could target Reddit, YouTube "sujets d'actualité TEF" compilations and Facebook groups to check the bank below.

How I counted.

• A prompt is one distinct ad or statement. Different ads for the same kind of thing, such as two different furniture-sale ads, are separate prompts.
• A sighting is one appearance in one source set. Each OPAL set (a combination page or a dated page), LFA, Hachette, FEH and Mocko each count once.
• OPAL pages that are exact re-posts were not counted again: EE combination 108 repeats 107, and EO combinations 101–104 repeat the four July 2024 sets.
• Because OPAL repeats prompts across its own combinations and FEH copies OPAL, sightings show what the circulating bank repeats. They are not proof of how often a prompt appears in live exams.

2. Current task definitions (verified)

Sources: EE sample, EO sample, 2016 manual pp. 30–33 (1 min "découverte du sujet" per section), examiner handbook Annexe 1, exam-day page (no drafts or notes in reading, listening or speaking; virtual keyboard for accents; no dictionary). The format has been unchanged since 2016, and the 2026 handbook shows the same durations (2026 livret, via tef-format-scoring.md). High.

• Writing is typed on a computer in LFA's software. If you finish Section A early you cannot go back to it. Scratch paper is collected at the end (tef-format-scoring.md §2.3, citing the 2025 livret du surveillant). High.
• No official maximum word count exists for either writing section. High (no official source states one).
• Instruction wording you will see:
  • EE-B: write a letter to the newspaper (200 words minimum) and develop at least 3 arguments.
  • EO-A: you read an ad for X, you are interested, and you call for more information; ask about ten questions.
  • EO-B: you read an ad for X and talk about it to a friend who [profile]; present the document and try to persuade them to [do X].
  • OPAL's prompts use the same templates. It adds a motive at the start of EO-A ("Vous recherchez une idée de sortie / un travail partiel / une activité artistique…"). Its EE simulation pages sometimes write "200 mots environ" (about 200 words) where the official text says "200 mots minimum".

3. What examiners reward

3.1 Official criteria

EE-A has four criteria (LFA Section A page, 30 Jun 2020). High.

1. Relevance of the information.
2. Quality of the information: development, details, illustrations.
3. Coherence and cohesion, plus the quality of sentences and vocabulary.
4. Spelling and punctuation.

LFA's advice for EE-A: don't copy, summarise or reword the opening; add new information; use the title for ideas.

EE-B has six criteria (LFA Section B page, 23 Jun 2020). High.

1. Relevance.
2. Quality of the arguments: developed, detailed, illustrated.
3. Internal coherence.
4. Cohesion.
5. Quality of sentences and vocabulary.
6. Spelling and punctuation.

LFA's advice for EE-B:

• A formal letter layout is not required and earns no points.
• Open with one short sentence.
• Skip long generic introductions and conclusions, because markers ignore them.

Both writing sections are also judged on these points (EE sample):

• whether the text fits the situation;
• how it is organised overall;
• how clear the information or argument is;
• whether you develop, nuance and specify your points;
• vocabulary, syntax (sentences, tenses, moods) and spelling.

Two independent markers grade each text on a CEFR grid.

EO (EO sample, examiner handbook pp. 8–10). High.

• Section A is marked on getting information, getting answers clarified, and holding a formal conversation. The examiner scores three things:
  1. how relevant your questions are;
  2. whether you take the initiative and handle the unexpected;
  3. the quality of the exchange.
• Section B is marked on presenting the situation, developing your ideas, and holding an informal conversation. The examiner scores three things:
  1. how you present the facts;
  2. the quality of your arguments;
  3. the quality of the exchange.
• Language is marked in both sections: vocabulary, syntax and pronunciation.
• The recording is marked again by a second LFA marker.

3.2 What the examiner is trained to do

All points from the 2024 examiner handbook. High.

Section A:

• Gives partial answers first (for example, "it depends, we have several packages"), so you have to follow up.
• Avoids asking you questions.
• Drops in surprises, such as special insurance, payment options or a current promotion, to test how you react.
• Ends the call politely at 5 minutes.

Section B:

• Tells you at the start that they will disagree.
• Uses a list of counter-arguments and common clichés.
• Steers you back if you forget to present the document.
• Pushes you to develop each argument and bring new ones, because candidates often repeat the same one.
• Avoids questions that make you tell a story instead of arguing.
• Ends at 10 minutes.

3.3 How many questions and arguments

4. What repeats (counted from the 301-prompt bank)

4.1 Clusters by section

EE-A: 54 prompts. 21 were seen 2+ times. Prompts per source family: OPAL 54, FEH 13, LFA 4, HAC 1.

EE-B: 49 prompts. 24 were seen 2+ times. Prompts per source family: OPAL 49, FEH 17, LFA 4, HAC 1.

EO-A: 108 prompts. 15 were seen 2+ times. Prompts per source family: OPAL 90, MCK 20, FEH 10, LFA 4, HAC 1.

EO-B: 90 prompts. 27 were seen 2+ times. Prompts per source family: OPAL 76, MCK 15, FEH 10, LFA 4, HAC 1.

4.2 The prompts seen most often

Each number is the count of separate source sets (OPAL sets, LFA, Hachette, FEH, Mocko) where the prompt appears. A prompt seen in OPAL and FEH counts 2.

• EE-A, 3 or more sightings:
  • 80-year-old vs car thieves (5; official);
  • sausage-seller row ends in prison (4);
  • garden "treasure" while planting a tree (3; official);
  • crêperie customers return with the police (3);
  • mice invade the opera (3).
• EE-B, 3 or more sightings (17 prompts):
  • package tours vs real discovery (6; official);
  • social networks are a permanent danger for children (5; official);
  • competition has made people selfish (5);
  • 4 sightings each: pointless to regulate teens' social media; computer classes for older people are useless; we must consume differently;
  • 3 sightings each: NGOs vs poverty; youth today vs earlier generations; no need to teach handwriting; machines replacing humans; free public transport; travel no longer only for the rich; films and series will replace books; a diploma does not guarantee a job; children should learn Chinese; 25–30-year-olds live with their parents out of comfort; cameras do not reduce crime.
• EO-A, 3 sightings: product tester job; e-bike rental or lease.
• EO-B, 3 sightings: volunteering at an animal shelter (friend loves animals); summer youth-group leader job; language lessons by phone; 22 km city race; escape room; charity-restaurant helper roles; multi-day cycling holiday.

4.3 Speaking scenario families (Sections A and B together)

These group different ads that share one scenario, counted with keyword rules over the 198 EO prompts. Families can overlap: an ad can fall into two.

What stands out:

• Castings for extras are the most repeated single scenario in EO-A: 6 different ads, one of them the official sample, plus 2 in EO-B.
• Second-hand buying and selling is almost only Section A: 9 of 10.
• Clean-ups, volunteering abroad and animals are mostly Section B.
• Language learning appears 3 times in Section B, including the official sample, and once in Section A, as a job ad.

4.4 The friend's profile in EO-B

Counted on the 77 Section B prompts from LFA, Hachette and OPAL; 4 could not be read.

• 39 of 73 (53%) give the friend a profile:
  • interest or trait (loves travel, cinema, gardening or reading; very fit; wants an extra activity; wants to get involved; cooks well; wants to learn a language; likes tech or photography; needs English; loves the outdoors; has the right qualities): 14;
  • free time: 5;
  • where they live or what they own (town, country house, suburb, flat, car): 5;
  • looking for work: 4;
  • holiday idea wanted: 3;
  • spare room or housing need: 3;
  • "might be interested": 3;
  • loves animals: 2.
• The remaining 34 just say "a friend".

Use the profile as your first argument when there is one.

4.5 Cross-checks with the theme-only sources

• Theme-only sources (LN, FFPR, TCO) list accidents, thefts, local events, discoveries and animal incidents for EE-A. Every one of these is a cluster in §4.1.
• For EE-B they list environment, technology, screens, work, city life, education and consumption. These are the top EE-B clusters, except work, which is small in this bank (2 prompts).
• frenchforpr says candidates report pet-care and environmental-volunteering themes in speaking. These match the Animals (8 sightings) and Clean-ups (8 sightings) families.
• All of this is undated, so low confidence.

5. Templates that cover the most prompts

All phrases are my own wording. Coverage = prompts in the bank that the template fits.

5.1 EE-A: three-paragraph news frame (covers 54 of 54)

1. What happened next (35–45 words). A time marker, then events in the passé composé with the scene in the imparfait. For example: Quelques minutes plus tard, … / Au moment où …, ….
2. Reactions and who stepped in (45–55 words). A witness quoted directly (« Je n'en croyais pas mes yeux », raconte …), then the people who acted: police, firefighters, zookeepers, doctors, staff.
3. Outcome and what comes next (35–45 words). The result (Heureusement, plus de peur que de mal.), the consequences (fine, arrest, record confirmed, family reunited), and the person's plans or lesson in the futur or conditionnel.

Add a title clue: every OPAL and LFA prompt has one, such as Une mamie en colère. Do not copy or retell the opening.

Vocabulary packs, in order of how many prompts they serve:

• animals on the loose: s'échapper, capturer, soigneur, vétérinaire, semer la panique;
• record or unusual journey: battre un record, exploit, parcourir, défi;
• odd everyday incident: insolite, se rendre compte que, finalement;
• windfall or find: gagner le gros lot, trouver un trésor, rendre l'argent, faire un don;
• missing person or reunion: disparaître, réapparaître, retrouvailles, ému aux larmes;
• rescue or emergency: porter secours, les pompiers, sain et sauf;
• crime or dispute: voleur, porter plainte, interpeller, condamné à;
• old message found: bouteille à la mer, retrouver l'auteur.

5.2 EE-B: 1+3+1+1 letter (covers 49 of 49)

1. One opening sentence that restates the claim and gives your position. For example: Vous affirmez que …; je ne partage pas entièrement cet avis.
2. Three arguments, each one claim, one explanation and one concrete example with a number or place. Rotate through these angles: individual benefit or harm; society and the economy; health; education and skills; environment; freedom of choice; safety; future generations.
3. One concession: Certes, … ; toutefois ….
4. A 1–2 sentence close with a solution or recommendation (Il faudrait plutôt …).

Why nuance works. By my count, 33 of the 49 statements are absolute or sweeping. They use words such as inutile, danger permanent, vont remplacer, il faut d'urgence, tous les pays, la plus grande menace, ne garantit pas, or make a blanket claim about a whole group. This count is my own judgement (low-medium). Accepting part of the idea while rejecting the absolute word gives three easy arguments (FEH gives the same advice). Five argument banks cover 78% of statements (38 of 49): technology and screens, society and values, environment and consumption, education, and travel.

5.3 EO-A: master question checklist (covers 108 of 108)

Opening: Bonjour madame/monsieur, je vous appelle au sujet de votre annonce … J'aurais quelques questions, si vous avez un moment.

Twelve questions that work for any ad:

1. What exactly is offered or included.
2. Dates, times and duration.
3. Address and how to get there.
4. Who it is for: age, level, group size.
5. Price.
6. What the price includes; discounts for students or groups.
7. How to pay; deposit.
8. How to sign up or book, and the deadline.
9. Equipment: provided or bring your own.
10. Safety and insurance.
11. Cancellation, refunds, weather.
12. Next step: visit, send a CV, call back.

Extra questions by ad type:

• Job or gig (16 prompts): tasks, hours, pay, contract length, experience needed, training, start date, interview.
• Course (14): teacher, class size, certificate, free trial.
• Buying second-hand (8): condition, age, reason for selling, price negotiable, delivery, viewing.
• Rental: minimum length, deposit, mileage, condition on return.
• Thrill or outdoor activity (19): minimum age or weight, fitness needed, instructors, photos or video.
• Service (17): free quote, how long it takes, guarantee, travel to your home.
• Outing or event (15): programme, tickets, reduced prices, food on site.
• Housing (4): rent and charges, furnished or not, lease, viewing.

Follow-up moves for partial answers and surprises (these score "initiative and handling the unexpected"):

• C'est-à-dire ?
• Qu'entendez-vous par … ?
• Et concrètement, ça coûte combien ?
• Si j'ai bien compris, … c'est bien ça ?
• Ah bon ? Et cette assurance, elle est obligatoire ?

Close: sum up, give your next step, thank them.

5.4 EO-B: present, persuade, handle objections (covers 90 of 90)

1. Hook (tu): Salut ! Écoute, je suis tombé(e) sur une annonce qui est faite pour toi.
2. Present the document in your own words (about 1 min): who, what, where, when, price, and 3 key features. If you skip this, the examiner sends you back to it.
3. Link it to the friend's profile (53% of prompts give one): Tu m'as dit que tu cherchais …
4. Eight angles; use 5–6: cost or value; health and well-being; meeting people or doing it together; skills, CV or career; fun, something new, memories; helping others or the planet; convenience and flexibility; low risk (free trial, beginners welcome, supervised).
5. Objection loop, repeated until time runs out. Acknowledge (Je comprends, mais …), answer with a detail from the ad, then offer a compromise (on y va ensemble, on partage les frais, on essaie une seule fois). Objections to expect:
   • too expensive;
   • no time;
   • scary or dangerous;
   • not good at it;
   • too tired;
   • too far;
   • bad weather;
   • not interested;
   • family commitments.
6. Close with a concrete plan (Je nous inscris pour samedi ?).

Language that helps: ça te dirait de …, et si on …, tu pourrais …, si tu viens, tu verras …, il faut absolument que tu …, même si ….

6. Study order by work-to-reward (EE and EO only)

7. Conflicts and open questions

1. Are the OPAL, FEH and Mocko prompts real?
   • None of the three names an exam sitting or a candidate.
   • OPAL includes all 16 official samples and Hachette's sample as "sujets d'actualité". Its ads have placeholder numbers, and some are visibly scanned from print.
   • FEH matches OPAL item for item.
   • Two possibilities: the live exam reuses a stable pool that includes the samples, or the banks pad themselves with samples. I could not test which.
2. Preparation time. Mocko says there is none. The official sources describe 1 unrecorded minute per section to discover the subject with the examiner, with no notes (2016 manual p. 32, handbook Annexe 1). Both are partly right: you get a minute, but you cannot prepare notes.
3. Wrong format descriptions on prep sites.
   • tefcanadaonline's speaking article describes a 3-part test with 1 minute of preparation per part. That is not the TEF format.
   • Mocko's "common writing topics" gives 200–250 and 380–450 word targets. Those do not match TEF either.
   • Ignore both.
4. Number of arguments in EE-B. The official minimum is 3; two sites say 2 (§3.3). Follow the official minimum.
5. "200 mots environ" vs "200 mots minimum". OPAL's simulations say "about 200"; the official text says "200 minimum". Write more than 200.
6. Dates. OPAL's only session-style labels are Jan–Mar 2024 and Sep 2024 for writing and 1–3 Feb 2024 for speaking. Everything else carries the month it was posted (Oct 2024 to May 2026). No bank gives month-by-month dated sittings for 2025–26 the way TCF banks do. Mocko also says TEF topics circulate as a recurring pool rather than as month-by-month lists.
7. Weighting of Section A vs Section B within writing and within speaking is not published (tef-format-scoring.md §4.1).

8. Full prompt table

Columns:

• Seen: number of separate source sets.
• Month(s): "(label)" is the date in OPAL's page title, which appears to name a session; other months are when OPAL posted the set; "official sample" is an LFA sample (2024 PDF; all 8 writing samples and 4 of the 8 speaking samples also appear in the 2016 manual); FEH pages date from 2026 and claim 2025–26 sittings; MCK is undated (2026).

Source codes:

• LFA: official samples.
• HAC: Hachette practice test.
• OPAL-EE101…113 = https://tefcanada.opal-ca.com/expression-ecrite-combinaison-NNN/ (with text copies at /ee-combinaison-NNNk/).
• OPAL-EEJ1/J2/J3/F1/M1/S1 = the dated EE pages /ee-combinaison-01-janv-2024/, -02-janv-2024/, -03-janv-2024/, -01-fev-2024/, -01-mars-2024/, -01-septembre-2024/.
• OPAL-EOF01/F02/F03 = /2160-2/, /2165-2/, /2201-2/ (labelled 1, 2 and 3 Feb 2024).
• OPAL-EOU4 = /2216-2/.
• OPAL-EO106…120 = /expression-orale-combinaison-NNN/.
• FEH and FEH-B15: the two French Exam Hub pages.
• MCK: Mocko.

EE-A (54 prompts)

EE-B (49 prompts)

EO-A (108 prompts)

EO-B (90 prompts)

Appendix B. Practice prompts the sites invented (not counted)

Use these for extra drilling. They follow the format, but the sites say they wrote them.

Method notes

• OPAL's speaking ads and some writing prompts are only available as images. I read them with Apple Vision OCR on images downloaded straight into memory; nothing was saved.
• OPAL page text and post dates were read in a browser pane, a few pages at a time, after a burst of parallel requests caused database errors on their server.
• The LFA and Hachette PDFs, and one test image, were cached automatically by the fetch tool. I extracted the PDF text where it was cached.
• Apart from small helper scripts in the session scratchpad (PDF text extraction, OCR and counting), the only file written is this report.