What AI tells prospective migrants about New Zealand

AI guidance about moving to New Zealand changes meaningfully with the prospective migrant's work, household and education circumstances. The most useful reviewed answers connect a place to an actual task: finding a technology employer, reaching a hospital, accessing dairy work, keeping hospitality rent manageable, or arranging safe accommodation before university. The weaker answers turn broad country advantages into confident claims about employment or visa access.

The October 2026 study collected 797 answers from four AI platforms. All 572 New Zealand-setting answers were collected. The remaining answers compare advice using other country settings. All 27 missing answers are from the China comparison on ChatGPT, Gemini and Google AI Overviews.

Of the 108 AI answers about where to live in New Zealand, 95 named Christchurch, 91 named Auckland and 70 named Wellington. The counts include every answer that names the place, whether the response is positive, neutral, negative or mixed.

The profile contrasts are more informative than the overall totals. Every one of the software case's 12 New Zealand answers about where to live names Auckland, Wellington and Christchurch. Every one of the dairy case's 12 names Waikato, Taranaki, Canterbury and Southland; none names Auckland or Wellington. The difference survives all three question wordings and all four platforms. Location counts.

The reviewed reasoning explains the contrast. Technology guidance emphasises employer concentration and networks. Dairy guidance follows farm locations and then considers schools, partner employment and travel to services. E10 E13.

Queenstown appears in all 12 hospitality location answers. In the reviewed Gemini answer, Queenstown is explicitly under “Regions to Approach with Caution”, while Christchurch, Palmerston North and Hamilton are presented as ways to combine ongoing trade with manageable housing costs. A simple name count would conceal the direction of the advice. E14.

The reviewed electrician preparation answer begins with professional assessment, then connects supervision, employers, visa rights and the household budget. The farm-family finance answer recognises that losing an employer-linked home can compound a job loss. Those are practical connections between circumstances and next steps. E28 E23.

Policy details are less dependable. Gemini's business-visa answer offers the closed Entrepreneur Work Visa alongside the newer Business Investor Work Visa. Google AI Mode explains the replacement in a dedicated visa answer, then names the old visa without the closure warning in a preparation answer for the same business case. Reliability therefore varies by question as well as by platform. E30 E32 E34.

All 12 New Zealand answers about where to live for the 17-year-old contain a homestay, pastoral-care, caregiver or Code-of-Practice term. The reviewed destination answer connects the school decision to later university access and approved accommodation. The same answer also wrongly suggests that the school automatically acts as the student's legal guardian. Other reviewed student answers mix secondary and tertiary financial thresholds or repeat an old general work-hours figure. E18 E29 E35.

In the seven-country comparison with complete four-platform coverage, Christchurch appears in 33–36 of 36 fixed location answers per country, and Auckland in 31–33. Individual answers still change their named places and explanation. The China dairy example retains the same four principal regions as the local example, while adding Asian-community and nearby-city detail. The answers differed in this round of research. The differences do not establish a lasting preference tied to nationality. E38 E39.

The report counts names and selected words across the full answer set. We read all 108 New Zealand answers about where to live to assess the tone toward each named place, and a separate selection of 41 answers explores practical advice across topics. A place-name count, the tone toward a place and a checked policy statement are different findings.

The destination answers do more than repeat a generic list of New Zealand's largest cities. Occupation and education stage change the locations that appear, while housing, access to services and the partner's circumstances change the reasons.

The comparison uses 108 answers: nine migrant profiles, three ways of asking where to live and four platforms, all using the New Zealand country setting. Each named city or region is counted once per answer. A city and its surrounding region remain separate labels.

Christchurch is the most frequently named city in the New Zealand comparison of places. The reviewed answers position Christchurch as a compromise between access to employers or institutions and the living-cost pressures attributed to Auckland. The cost comparisons are AI claims rather than a verified rental-market ranking.

Of the 95 answers naming Christchurch, 60 describe benefits without drawbacks, 32 give both benefits and drawbacks, two describe drawbacks without benefits and one mentions the city without an evaluative view. Auckland is named in 91 answers: six are positive, 68 mixed, 15 negative and two neutral. Wellington is named in 70 answers: 15 are positive, 41 mixed, 13 negative and one neutral.

The tone review covers all 108 New Zealand answers about where to live, with one label for each named place in an answer. Mixed answers can recommend a place while also describing costs, weather or employment limits. Neutral answers can name a place only in an institution or source title. The complete place-by-place review preserves the supporting wording and reason for each label. The labels describe AI wording, not verified destination quality.

Every place listed in the profile table appears across all three question versions for that profile, although the number of answers and platforms naming each place varies. The complete profile table and platform-by-wording table retain the less frequently named places and zero counts.

Software and data professional. Auckland, Wellington and Christchurch form a consistent core. In the reviewed Gemini answer, Auckland is associated with commercial employers, Wellington with government and public-sector data, and Christchurch with engineering, agritech and software. The question about the applicant's specialisation matters: a large commercial data team and a scientific computing employer imply different searches. The answer's general enthusiasm about tech opportunities does not establish how quickly an applicant without work rights would find a job. E10.

Nurse moving with a family. The destination reasoning includes the partner's work and children's schooling, rather than treating a nursing vacancy as the only criterion. Gemini contrasts major hospitals, regional base hospitals and rural services. The answer explicitly asks about the partner's employment goals. The broad claim that nursing demand is high nationwide needs more qualification than the answer provides; the section on job prospects and policy review show the conflicting evidence. E11 E21.

Electrician moving with a partner. Auckland and Waikato each appear in 11 of 12 New Zealand answers about where to live. The reviewed Gemini answer gives Christchurch a manufacturing and industrial-work rationale, and Waikato a mixture of residential, agricultural and industrial demand. Registration and supervision are central practical conditions. Promises that sponsorship is straightforward are not established by naming growth industries. E12.

Dairy worker moving with a young family. Waikato, Taranaki, Canterbury and Southland each appear in all 12 answers. The reviewed guidance distinguishes farm scale and climate, then considers how far the household would be from schools, medical services and alternative work for the partner. Hamilton or Christchurch can appear as nearby service centres without being the suggested farm location. The four regions provide the clearest contrast with the technology professional's city-focused answers. E13 E38.

Hotel or restaurant manager. The reviewed Gemini answer prioritises demand spread across local customers, business visitors, institutions and tourism. Christchurch, Palmerston North and Hamilton are suggested on those grounds. Queenstown appears as a rental-cost and seasonality caution. The answer even says institutional demand “guarantees” year-round trading in one regional discussion, an assurance the capture does not substantiate. The useful reasoning is the distinction between tourism appeal and a financially sustainable household, not a promise that a particular market guarantees work. E14.

Postgraduate student with a limited budget. The reviewed Gemini location answer connects Canterbury, Otago and Waikato to institutions and fields, then contrasts the living-cost pressure attributed to Auckland and Wellington. The answer asks for the intended field before narrowing the choice. The model's assertion that targeting a shortage field drastically shortens the job search is a stronger claim than the evidence supplied. E15.

Small-business owner. The destination choice remains conditional because the case does not name a sector or acquisition target. The reviewed answer contrasts metropolitan customer markets with primary-sector supply chains and tourism markets. The useful clarification is whether the business needs a large local consumer base, business customers, visitors or export links. Neither a regional economic profile nor the ability to buy a business establishes immigration eligibility. E16 E07.

Partner of a New Zealand citizen with a preschool child. The reviewed location answer considers work, childcare, housing and commuting. Christchurch is framed around space and family amenities; Auckland around a larger labour market; Wellington around particular occupations; Hamilton and Tauranga around different cost and lifestyle trade-offs. The answer asks for the applicant's work background, a fact absent from the case. A family-oriented answer is therefore more defensible than a single supposedly optimal city. E17.

Seventeen-year-old international student. Auckland, Christchurch, Wellington and Dunedin recur across school-to-university guidance. The reviewed answer connects cities to university options, transport and support, while noting that a trusted caregiver could determine the region. The best question for the student becomes which school can provide an appropriate academic pathway and approved care arrangement in a suitable location. The legal-guardian claim in the same answer remains a material correction. E18.

The contrast between the tech and dairy cases is occupational. The contrast between the two student cases is partly educational and partly about age and support. The contrast between a nurse and the nurse's whole household is about the partner and children as well as the applicant's work.

A publication that reduced the answers to one “best place for migrants” list would discard those distinctions. A useful presentation retains the reason, the condition and the stated trade-off beside each destination. Counting first choices and positive recommendations would require reading and assessing every answer about where to live. The place-name tables do not provide those measures.

The reviewed suitability answers do not give every profile the same rationale. Gemini frames the software professional's decision as a choice between lifestyle priorities and career or compensation scale. The hospitality manager receives an explicit “Yes, conditionally”, linked to pay and rent. The postgraduate receives a stronger financial test: the proposed qualification should serve a specific career plan and the budget must work. E01 E05 E06.

For the nurse, the reviewed answer advises starting registration before relocation and securing employer support. For the electrician, the answer explains a period of supervised work and local assessment before independent practice. Those transition periods affect both suitability and savings. The encouraging language elsewhere in the answers should be read alongside those conditions. E02 E03.

The report does not assign an overall recommendation-strength score to the four platforms. The full-answer review supports the stated examples; a valid comparative score would need consistent interpretation of every eligible suitability answer.

Among the 72 drawbacks answers, 55 contain a housing-cost term and 53 contain a loneliness, isolation, homesickness, belonging or support-network term. The counts show which words appeared, not how serious each drawback is.

The reviewed software drawbacks answer goes beyond the job market to describe social circles, nightlife and dating. Some observations may help a single newcomer think about support; other statements are broad cultural assumptions unsupported by the person's facts. The model moves from a work situation into a wider account of what the applicant might value. E19.

For the dairy family, the reviewed drawbacks answer connects long seasonal hours, rural isolation and housing tied to employment. The same case's suitability answer describes on-farm accommodation as a major benefit. The two answers expose different sides of the same arrangement: an easier arrival can also increase dependence on one employer. E04 E20.

Nursing shows why a dedicated question about practical prospects matters. Gemini's suitability answer says nurses are heavily in demand. Google AI Overviews' answer about job prospects instead notes lower general demand and particular specialist exceptions. The latter answer still becomes very confident about competitiveness after registration, so recognising one constraint does not validate the whole answer. E02 E21.

The nursing regulator's own guidance encourages prospective applicants to explore employment before registration or travel. Qualification recognition, legal permission to practise and an actual job offer should remain separate facts in a migrant's plan. Nursing Council guidance.

The postgraduate question about practical prospects reveals a design problem as well as a model response. The exact question states an undergraduate degree, a limited budget and a career goal, but omits New Zealand and international-student status. Google AI Mode begins with domestic fees and suggests combining work with part-time study. The answer cannot be assessed as though the missing international context was supplied. Future wording should preserve the international applicant's status while keeping the natural voice. E22.

All nine profiles were asked about money in two ways on four platforms, giving 72 answers. The table shows how often the listed words or topics appeared.

The useful finding is the breadth of planning topics, not the accuracy of every price. The farm-family answer describes vehicle costs, the partner's restricted employment options, the gap before the first pay and the possibility of needing to move house when employment ends. Those are specific ways a household budget can fail. E23.

The postgraduate finance example separates fees, living costs, exchange rates and visa delays, then warns against relying on uncertain part-time work. The same answer provides detailed fee and rental ranges without establishing that the figures are current quotes for an actual provider or property. A detailed budget template can be useful even when the inserted prices need replacing. E24.

For the younger student, the distinction between school and tertiary living-funds requirements is consequential. The verified benchmarks are NZ$17,000 annually for school years 1–13 and NZ$20,000 for tertiary and other covered study. Tuition and outward-travel funding are additional. A visa funds benchmark does not estimate the full cost of the proposed life. INZ student fund requirements.

The 40 settlement answers apply to the five cases with defined family or learner-support needs. Thirty-one contain a transport or driving term, 31 contain a childcare or school term, and 27 contain a loneliness or support-network term. The other cases were deliberately not given separate settlement questions.

The reviewed nurse answer describes the simultaneous tasks of registration, renting, school entry and building support without nearby family. The citizen-partner household receives an explanation of childcare waiting lists, extra hours and the strain of job hunting while caring for a preschooler. Those observations recognise that access in principle and practical availability are different. The benefits named in settlement answers still depend on eligibility and local availability. E25 E26.

The 17-year-old's settlement question produces a counterexample. The question says the student will live away from parents but does not identify an international student whose parents remain overseas. Google AI Mode discusses StudyLink. The resulting answer is poorly suited to the intended case, but the missing question facts are part of the explanation. E27.

The reviewed electrician preparation answer places assessment before relocation and links the assessment outcome to supervision, employer selection and savings. The reviewed student preparation answer puts accommodation, the offer of place and academic progression near the start. The same student answer then gives an overgeneralised first-year funds figure. E28 E29.

Immigration New Zealand or INZ is explicitly named in 53 of 72 preparation answers. Naming an authority is useful for orientation, but naming an authority does not establish that every adjacent claim is supported by the authority's current guidance.

All four platforms answered the same 143 questions using the New Zealand country setting, allowing a like-for-like comparison.

The table counts sources attached to an answer separately from web addresses in the answer text. Each recorded website counts once per answer. The web-address count includes image and file links. A separate downloadable table excludes image and file links and therefore reports 65 rather than 68 Google AI Overviews answers.

Gemini has sources attached to 122 of the 143 New Zealand answers, even though none of the saved answer text contains a web address beginning with http:// or https://. A recorded source does not prove that the page supports every statement or explains why AI gave the answer. An answer without a link may still have used web search or shown separate citations in the live interface.

Source records were retrieved on 26 October 2026 for the same 572 New Zealand answers. No new answers were collected. The original collection dates and the 22 October policy checks remain unchanged. Answers with sources and links · Websites recorded for each answer.

The median is the middle answer length after sorting from shortest to longest. ChatGPT's median answer is substantially longer than the other platforms' medians. More text creates room for explanations and conditions, but length is not an accuracy measure. Google AI Overviews also has a longer median answer than Google AI Mode in the study.

The visa questions deliberately omit citizenship and ask what additional information is needed. Citizenship, nationality or passport words appear in 39 of the 72 answers.

The table counts words, not whether the platform correctly handled every rule that depends on nationality or asked for the applicant's citizenship. The reviewed AI Mode visa answer for the younger student asks for the passport country. The reviewed Gemini business answer instead asks about capital, industry and timing. Both sets of facts can matter, but the country setting used to collect an answer should not be treated as the applicant's citizenship. Words and phrases counted.

The complete comparison uses the same nine questions about where to live on all four platforms, with New Zealand, India, the Philippines, the United Kingdom, South Africa, Fiji and Australia as country settings. Each country has 36 answers.

The broad urban and regional names remain familiar across the seven collection settings. Individual place-name sets are less stable: only zero to three of each comparison country's 36 answer pairs have exactly the same set as their matching New Zealand answer. The difference can be an extra nearby town, a caution or a service centre; the difference is not necessarily a changed recommended destination.

The country comparison table pairs each answer with the New Zealand answer to the same question on the same platform and lists the places added or omitted. Readers can inspect the full answers before deciding what a difference means.

Only Google AI Mode returned the nine China answers about where to live. The separate eight-country table therefore compares the same nine questions on Google AI Mode alone. The nine China answers are not compared with 36-answer totals from other countries.

In the reviewed dairy comparison, the New Zealand and China answers both open with Waikato, Taranaki, Canterbury and Southland. The China answer adds nearby towns and references Asian and migrant communities. The younger-student China answer adds Asian amenities and international flight access. The case text does not establish an Asian nationality, so the added content is an observed framing choice rather than evidence about a real applicant's identity. E38 E39 E40.

The software professional's three main cities and the dairy family's four principal regions appear across all three question versions on all four platforms. Less frequent places are less consistent: Queenstown appears in only two of the software professional's twelve answers about where to live, both using the same question version. A place appearing under only one question version should not be presented as a consistent recommendation for the profile. Results by question version.

The task can also change the answer. Google AI Mode identifies the replacement business visa in both versions of the visa question, but its preparation answer names the old route without the same warning. Another round using the same questions and settings would be needed to establish whether the difference recurs. E32 E34.

The reviewed answers contain useful practical guidance and consequential mistakes in the same response. The evidence does not support naming one uniformly reliable platform.

The six original policy checks below were made on 22 October 2026; the review was expanded on 28 October as described below. The checks concern selected claims; the checks are not a random sample or an accuracy rate for all 797 answers.

Gemini's dedicated business-pathway table offers an Entrepreneur Work Visa at a NZ$100,000 investment threshold alongside the Business Investor Work Visa. The question comes from a prospective migrant considering a new move, so the omitted closure warning is consequential. INZ announced the closure of the Entrepreneur Work Visa in August 2025. The current Business Investor route starts at a NZ$1 million qualifying investment and requires separate reserve funds. The difference changes which plans are even worth investigating. E30. INZ closure announcement, current Business Investor requirements.

Counterexamples matter. ChatGPT explicitly says fresh Entrepreneur Work Visa applications are closed. Google AI Overviews also states the closure. Google AI Mode correctly describes the replacement in the visa question, but its preparation answer lists the Entrepreneur Work Visa as an option to review without the closure qualification. Even current knowledge in one answer does not ensure the next answer applies the same distinction. E31 E33 E32 E34.

The reviewed Gemini nurse answers describe heavy or nationwide demand. The Nursing Council's current guidance says recent demand for internationally qualified nurses is lower outside particular specialities and encourages checking employment before registration or travel. A registration pathway and a place on an immigration list do not demonstrate that an appropriate employer is recruiting the applicant now. E02 E11. Nursing Council employment guidance.

Google AI Overviews recognises lower general demand in the dedicated answer about practical prospects. The same answer later becomes more confident about the applicant's competitiveness after registration. The more accurate constraint should be retained without adopting every subsequent assurance. E21.

Gemini's younger-student visa answer says tertiary student visas generally allow 20 hours of work per week. Current INZ guidance allows eligible students up to 25 hours, subject to their own visa conditions; existing visa conditions and research-degree rules can differ. The answer gives an old general figure without explaining the current distinction. E35. INZ work conditions.

Google AI Mode's preparation answer for the 17-year-old quotes about NZ$20,000–22,000 for the first year. The school-level living-funds benchmark is NZ$17,000 a year; the tertiary benchmark is NZ$20,000. The Google AI Overviews visa answer correctly distinguishes the two amounts. An adequate real budget can be higher than either benchmark, but a planning estimate and an official requirement should be labelled differently. E29 E37. INZ funds requirements.

The reviewed Gemini destination answer says the high school will act as the 17-year-old's legal guardian. INZ's provider guidance requires confirmation that accommodation for students aged 17 or younger complies with the learner-care Code. INZ separately treats legal guardianship as a legally recognised status supported by evidence. The answer conflates the provider's pastoral-care duties with legal guardianship. E18. Provider accommodation requirements, INZ legal-guardian evidence.

Gemini's suitability answer says the applicant can transition to a partner resident visa after living together in New Zealand for 12 months. The published partnership requirements concern evidence of living together for the relevant period; the period is not restricted to time in New Zealand. Gemini's search-style pathway answer for the same case describes applying while still overseas. E08 E36. INZ partnership evidence.

The postgraduate question does not name New Zealand or say the student is international. The younger student's settlement question omits international status and parents remaining overseas. Domestic tuition and StudyLink guidance appear in the resulting AI Mode answers. StudyLink has residency and course requirements, but the omission in the question prevents a clean test of whether the platform recognises the intended international case. E22 E27. StudyLink eligibility.

The original answers show how a natural question can be interpreted when important details are missing. Both questions should be revised before a future study. The original wording remains visible, and the resulting advice is not counted as a straightforward AI error.

The policy and context ledger preserves the checked claims, exact source answers, primary sources, dates and assessment.

The expanded review checks 47 statements across 27 policy questions, using 30 complete saved answers covering all nine migrant profiles and four platforms. The verdicts are 28 correct, nine wrong, four outdated and six unclear. The six original decisions remain visible with their original review dates. The selection includes useful factual guidance as well as problems; the selected statements do not give a random-sample accuracy rate or certify every statement in the reviewed answers.

Correct examples include the closed Entrepreneur Work Visa, the separate school and tertiary living-funds amounts, partnership applications from overseas, accredited-employer job requirements, pathway-study options, business-investor reserves and visitor maintenance funds. The wider review also finds false points categories for skilled residence, a future Graduate Diploma work route presented as already available, an outdated overseas-car-licence limit and an incorrect claim that wages cannot be paid without an IRD number. Each verdict and exact passage appears in the statement review.

Both visa-question styles were read for every profile. Thirteen of the 27 policy questions have checked assertions from only one wording; those checks do not establish consistency across wordings. The measure coverage table records the precise questions, answers and styles behind each measure. The statement download retains every original answer reference, exact quote, verdict, dated official source and explanation.

The checks use rules applying during the 15–20 October collection. For example, the wider Graduate Diploma post-study work eligibility starts on 16 November 2026, while the overseas-car-licence guidance distinguished the October 18-month limit from the announced November change. Missing international status and other prompt facts remain interpretation limits, not scores for isolated model error.

Prospective migrants can use AI to develop questions, compare broad options and build a preparation checklist. The most useful reviewed answers connect decisions that are easy to consider separately: professional registration with time out of work, rural jobs with the partner's employment, housing with dependence on one employer, and school choice with care and later university entry.

The policy examples show why a plausible plan still needs checking where permission, funding and timing depend on the rules. A named visa can be closed, a stage of study can change the funds required, and a family's eligibility can depend on evidence the question did not mention.

Public agencies, employers and education providers can help by explaining the steps together: who qualifies, what remains uncertain, which costs are estimates, and which decision should come first. A list of national advantages is less useful than practical guidance for a nurse's household, an electrician working under supervision or a student aged 17.

Regional organisations should explain why a place might suit a particular occupation and household. Christchurch's frequent appearance does not mean every migrant should move there. Waikato appears for different reasons in dairy and technology answers. Queenstown's appearance in hospitality answers can reflect a warning about costs.

The study uses nine migrant profiles and eight topics: whether to move, drawbacks, visa options, places to live, work and study, money, settling in, and preparation.

Researchers wrote the people, circumstances and priorities. The profiles test how advice changes when relevant facts change; the profiles do not measure the size or preferences of real migrant groups. The study does not follow households over time.

The study contains 143 questions. Every question was asked on ChatGPT, Gemini, Google AI Mode and Google AI Overviews. Each topic was asked in at least two ways: a fuller question using the same request across profiles and a shorter, search-style question. Questions about where to live also have a conversational version. Work, study and business questions apply to eight profiles; settling-in questions apply to five.

We planned 824 answers, counting each question, platform and country setting separately. Every question was asked using the New Zealand setting. Only nine questions about where to live were also asked using all eight country settings. The country setting does not identify the applicant's citizenship.

The original collection began on 15 October 2026 New Zealand time and finished on 20 October. The available answers were downloaded on 21 October.

The archive contains 797 different, non-empty answers. The saved questions match the approved wording, and no duplicate or unexpected answers were found. All 27 missing answers are from the China questions on ChatGPT, Gemini and Google AI Overviews. Missing answers are not counted as advice against a place.

The complete answer set contains about 604,000 words. We counted place names, selected words and phrases, recorded sources, web links and answer lengths. The counting rules explain each measure, and the supporting tables retain the results.

We read 41 selected answers in full to examine the advice and reasons. The reviewed examples cover all eight topics, all nine profiles, all four platforms and several country settings. The answers were chosen for useful contrasts, not at random. The findings therefore do not give an overall error rate or a complete ranking of recommendations.

We also read all 108 New Zealand answers about where to live and assessed the tone toward each of the ten places in the comparison. The review covers 630 named place-and-answer combinations. Positive means benefits without drawbacks; negative means drawbacks without benefits; mixed means both; neutral means a factual or incidental mention. A name found only in an institution or source title is neutral unless the answer also evaluates the place. General visa and job-search cautions do not change the label for every place. Each place receives one label per answer; the original name counts remain unchanged.

The coloured bars divide answers naming a place by tone. The labels describe AI wording, not verified destination quality or first-choice recommendations. The review was completed on 27 October 2026 using the original answers. The place-by-place review retains every label, exact supporting passages and the reason for the decision.

A place name can appear in a recommendation, warning, source label or comparison. The destination tables count appearances, not endorsements. Each website recorded as a source counts once per answer. Links in answer text are counted separately and include image and file links. A recorded source does not prove that the page supports every statement or explains why AI gave the answer.

Direct platform comparisons use the same 143 New Zealand questions. Comparisons across all four platforms use the seven countries with complete results. The eight-country comparison uses Google AI Mode alone. Differences in one round of research do not establish that geography caused the differences or that the answers will stay the same.

Two selected student examples illustrate missing details in the questions; the complete question-context review below extends that limitation across the full catalogue. The findings describe the missing details rather than treating the resulting advice as a straightforward AI error. The original questions remain visible and unchanged; future questions should restore the destination and international-student details.

Selected policy statements were checked against Immigration New Zealand, the Nursing Council and other relevant official sources on 22 October 2026. The checks distinguish incorrect advice from questions that left important details unclear.

The original AI answers remain unchanged. Full answers, questions, missing-answer records, country comparisons, source records and counting tables are available from the evidence appendix.

Sources were retrieved on 26 October 2026 for the same 572 New Zealand answers. The source update did not collect new answers or change the questions, place counts or 22 October policy-check date.

A complete question-by-question review corrected fact descriptions for 57 of 143 questions. Twenty-one questions do not name New Zealand; other shorter questions omit qualifications, family details or funding arrangements from the intended profile. Student questions do not explicitly establish international fee status. We interpret the original answers using only the facts actually supplied. The review does not change the original wording, answers or descriptive counts. See the question context review and metadata review record.

The source wording table reports both answer totals and question styles, with each original answer reference. For the nurse’s visa questions, nursingcouncil.org.nz appears in three of eight answers, all from fixed-structure questions: one of two styles. Source repetition within one style does not establish consistency across styles. The original 1,462 answer/website associations are unchanged.