Decode Your Health with AI

Give your health information a home.

I recommend keeping a folder on your computer for your health records and what you learn. With ChatGPT Work, you can give ChatGPT access to that folder and ask it to create and update the documents for you.

1. Open the ChatGPT desktop app and sign in. Create a folder on your computer called My Health.
2. Create a local project in ChatGPT with the same name and attach that folder. In the project menu, choose Edit project, then Add folder.
3. Select Work and Work locally. Start a conversation called Health overview.
4. Copy the prompt below, paste it into that conversation, and answer the questions. ChatGPT will create the documents. You do not need to write or format them yourself.

Help me organize my health information. First check that you can read and write in the My Health folder I attached. If you cannot, tell me what access is missing. Do not claim to have saved files without checking.

Inspect any existing files and preserve them. For a new folder, create master-health.md for my overview, records/originals/ for original reports, and topics/ for research. Create the documents yourself; I do not have templates. Never delete or overwrite original reports. Ask before reorganizing existing work.

Ask me a few questions at a time about my main concerns, diagnosed conditions, medications and supplements, allergies, operations, family history, sleep, food, activity, stress, mental health, work and environment, and what I want to understand. Wait for my answers and let me skip anything. Take a holistic view of my health rather than reducing it to lab values.

Build a concise master-health.md with what I tell you: my profile, a health timeline, current care and medications, lifestyle context, key questions, and a section for prevention and follow-up discussions. Keep documented medical facts, my observations, and unanswered questions separate. Mark missing information as unknown. Do not invent a diagnosis or add research suggestions to my current medication list.

Include an index of my topic conversations. Each entry should have the exact conversation title I provide, its saved topic file, a short summary, and the last update. Do not invent conversation links or publish my information.

Show me the overview so I can correct it. Tell me which files you created and where to find them.

Read the overview it creates and correct anything wrong. It is fine to leave things unknown or skip a question.

Give it the context a report cannot provide: what your days look like, your eating and sleeping habits, movement, stress, mental health, family history, and what you want to improve. Add medications and supplements with their actual names and doses. Include relevant operations, hospital stays, and earlier treatments, even when the paperwork is incomplete.

ChatGPT Work is currently available on Free. Paid plans give you more usage for longer research. They still have limits. Check current plans and allowances.

If the folder options are missing, check the project instructions and Work setup guide. A regular web project uses uploaded or connected files; it does not have direct access to your computer folder.

Add the records you have.

Start with a small batch: a blood test, a consultation report, a prescription, or an imaging report. Put them in the records/originals folder ChatGPT created. A readable scan of a paper report can work too.

Over time, include specialist letters, surgical reports, discharge summaries, allergy assessments, and older results. Ask ChatGPT to keep a dated chronology rather than replacing old results with the latest number. A result within the reference range can still be useful context for a later comparison.

Return to Health overview and use this prompt. Replace the text in brackets with your filenames.

I added these reports to records/originals/: [filenames].

Read master-health.md and the named reports. Tell me which files and pages you could actually read. Extract every reported clinical value and relevant narrative into records/test-results.md, organized by year and date, preserving earlier entries and avoiding duplicates. Do not discard results because they are within the reference range. Work in batches if needed and list anything not yet processed.

Keep the report date, source filename and page. For test results, include the exact value, units, reference range, and the report's flag. Keep the clinician's comments separate from your interpretation. Mark new extractions as not yet checked by me.

Do not guess unclear text or missing values. List anything unreadable, inconsistent, or uncertain. Explain unfamiliar terms in plain language.

Show me the extraction and ask me to compare important details with the originals, especially numbers, medications, doses, and allergies. Wait for my corrections before adding confirmed information to master-health.md. Preserve unrelated content and tell me which topic summaries might need updating.

Keep the originals unchanged. Do not diagnose me or recommend treatment changes from a result alone.

Compare the extraction with the original reports. Check dates, numbers, units, medication names, doses, and allergies. Ask it to flag anything it cannot read instead of guessing.

You can also tell it what is missing from the records: symptoms, habits, past events, or something you want to understand. Keep your observations distinct from what a clinician has documented.

No reports yet? Start with what you know and continue to your first topic. You can add records later.

One conversation for each topic.

Inside My Health, start a new conversation and name it after the topic you want to explore. Keep your questions about that subject there. You do not need to create every conversation now.

A condition you have

Understand the diagnosis, your reports, and the questions you have about your care.

Allergies

Keep confirmed allergies separate from reactions or triggers you are still investigating.

Food

Explore your eating habits and nutrition questions, with your health history in view.

Sleep

Bring your sleep routine, observations, and questions about what might affect your nights.

Digestion

Keep symptoms, relevant meals, tests, and questions together over time.

Back pain

Follow the history of the pain, consultation reports, and advice you have received.

Copy this prompt into your topic conversation. Replace [TOPIC] with its name. Ask for thorough research, then keep asking about the parts you do not understand.

This conversation is about [TOPIC]. Help me understand it thoroughly, one question at a time.

Read master-health.md, any existing file for this topic, and relevant records. Tell me what you read. Ask about missing context before drawing personal conclusions.

Do thorough research using current clinical guidelines, systematic reviews, and relevant original studies. Link the sources you actually consulted and give their dates. Explain unfamiliar terms, what the evidence supports, conflicting findings, and what remains uncertain. If you cannot browse or verify a source, say so.

Take a holistic approach: consider nutrition, sleep, movement, stress, mental health, relationships, environment, and relevant microbiome research alongside conventional care. Investigate complementary and alternative approaches, including functional medicine, traditional Chinese medicine, or Ayurveda when relevant. Assess the specific intervention and outcome, not just the label.

Challenge assumptions in mainstream guidelines and official studies as well as alternative-health claims. Examine study design, sample size, comparison group, absolute effects, duration, replication, applicability to me, and funding or conflicts of interest. Seek evidence against your initial conclusion. Do not assume funding invalidates a finding, or that a traditional or natural approach is effective or safe. Distinguish missing evidence from evidence of no benefit.

Include a dated timeline, possible connections across topics, conventional and complementary options, questions about tests or specialists, and practical lifestyle ideas. Separate generally low-risk habits from interventions needing professional review. For supplements, assess whether any are justified, possible interactions, and what should be checked with a clinician or pharmacist; do not produce an automatic supplement stack.

Keep facts from my records separate from my observations and research hypotheses. Link medical facts to the source file and page. Do not diagnose me, invent citations, prescribe individualized treatment, or suggest delaying effective care. Make clear which options are research questions rather than an agreed care plan.

Save the explanation, sources, date, and questions for my clinician in a file under topics/. Use a clear name such as sleep.md. Add or refresh this topic's entry in master-health.md with the conversation title, file, short summary, and update date. Read the current master before saving and preserve other topics. Ask me for the exact conversation title if you cannot see it; never invent a link.

Show me what you saved and what I should check. If you cannot finish, save what is complete and list what remains.

For example, in Sleep you might describe waking during the night, explain your routine, and ask what the research says. Ask it to distinguish general evidence from what your own information can establish. A week later, bring new observations back to Sleep.

Open the sources behind important claims. Ask which ones are current, where the evidence disagrees, and what remains unanswered. The aim is to understand your situation well enough to ask informed questions, not to treat an AI answer as a diagnosis.

The topic prompt also asks for holistic and alternative perspectives, criticism of the studies, and practical questions about prevention, lifestyle, and treatment options. Use the broader research prompt below when you want to push an existing conversation further.

Come back to the same conversation.

When something changes, reopen the relevant topic. Add a new report, describe what happened, or ask another question. You can talk normally; this prompt helps when you want the saved information updated too.

Here is something new about this topic: [my question, observations, or new report filename].

Read the current master-health.md and this topic's saved file before answering. Tell me what you read. For a new report, check the exact source, date, values, units, and reference ranges; flag anything unclear for my verification.

Explain what this information changes and what it does not establish. Consult and link current reliable sources if more research is needed. Otherwise say how old the existing research is.

Update the topic file with the date, source references, and a short note of what changed. Keep my observations separate from documented findings and possible explanations. Read the master again and refresh this topic's index entry without changing unrelated topics. Keep conflicts visible and ask me to resolve them.

Tell me what needs checking and what is worth asking my clinician. Do not change my treatment plan or treat an AI suggestion as a medical fact.

The prompts ask ChatGPT to save the useful parts and update the master. Let one update finish before starting another. Files do not update on their own, and separate conversations should not be assumed to remember each other. Ask each one to read the current master and relevant topic files.

Keep a backup of your health folder somewhere you trust. If a conversation gets too long, ask it to save the current summary and unfinished questions. Start a continuation, ask it to read those files, and update the conversation title in the master index.

Ask for a holistic view. Question the studies too.

I find Codex very cautious on medical questions, with a strong leaning toward established medical guidance. That is my impression of using it, not a guarantee about every answer. I would explicitly ask it to look more broadly rather than stop at the first conventional explanation.

For me, a holistic approach means looking at the person behind the records: food, sleep, movement, stress, mental health, relationships, and the environment they live in. Ask how these might interact with a condition or treatment. Explore relevant microbiome research without assuming every symptom comes from the gut.

You can ask it to investigate functional medicine, traditional Chinese medicine, Ayurveda, acupuncture, mind-body practices, and other complementary or alternative approaches. Be specific about the intervention and the problem you want to address. The evidence for one practice does not validate every claim made under the same label.

Challenge an official study or guideline by examining how its conclusion was reached. Ask about the comparison group, sample size, absolute benefit, follow-up period, replication, and whether the participants resemble you. Look at funding and conflicts of interest, but do not use them as automatic proof that a result is wrong. Apply the same scrutiny to an alternative-health practitioner selling a test or supplement.

Keep the distinction clear: a complementary approach is used alongside conventional care; an alternative is used in its place. Researching an option does not mean replacing an effective treatment with it. Herbs and supplements can have side effects and medication interactions. NCCIH's guide to evaluating complementary approaches is a useful starting point.

Revisit your analysis of this topic with a holistic approach. Read the current master and topic files first.

Look beyond medication and individual lab values. Consider nutrition, activity, sleep, stress, mental health, relationships, relevant environmental exposures, and microbiome research. Include specific complementary or alternative approaches relevant to my question, without assuming an entire tradition works or does not work.

Critically review the mainstream recommendation and its strongest credible challengers. For key studies, compare design, sample size, control group, absolute rather than only relative effects, follow-up, replication, funding, and relevance to someone like me. Separate clinical outcomes from changes in a biomarker. Name the actual methodological weaknesses; do not imply fraud without evidence.

Use the same standards for functional medicine, traditional Chinese medicine, Ayurveda, supplements, and other alternative claims. Label findings as well supported, preliminary, uncertain, or contradicted by evidence. A lack of trials is not proof of either benefit or harm.

Explain plausible benefits, risks, costs, medication interactions, and the risk of delaying effective treatment. Include practical lifestyle options and questions for appropriately qualified practitioners. Say what evidence would change your conclusion.

Link sources you actually read, including important disagreement. If you cannot browse, say so. Update this topic's research file and its master-index entry, preserving existing facts and separating hypotheses from a plan I have agreed to.

Bring back the questions that connect the whole picture.

These were an important part of the original guide. Use them in Health overview or the relevant topic conversation, with the research prompt above:

• Timeline and trends: what changed before a symptom appeared, after a treatment, or between two tests? Ask it to separate coincidence from a supported explanation.
• Blind spots: what information is missing, which assumptions are driving the answer, and what credible evidence might contradict it?
• Treatment options: what has already been tried, what helped or caused problems, and which established or experimental approaches are worth discussing? Keep clinical trials distinct from proven care.
• The right help: which questions belong with a primary-care clinician, a specialist, a dietitian, a pharmacist, or a qualified complementary practitioner? Ask what each could contribute.
• Supplements: is there a documented need or a supported benefit for this situation? Ask about harms, interactions, product quality, and cost. A useful answer may be that no supplement is justified.
• Daily life: turn vague advice into realistic habits you can discuss and follow. Record what you tried, how you felt, and what changed without assuming you have proved a cause.

Keep a short list of priorities: what you can learn or organize now, what to discuss at the next appointment, and what needs professional assessment before acting. Keep proposed changes separate from your current care plan.

Prevention is key.

I would make a prevention review part of this process, including a discussion of blood testing. Do not wait for a crisis to become interested in your health. Build a picture of where you are now, identify risks worth addressing, and decide what is useful to follow over time.

Start a Prevention conversation. Bring your family history, existing results, symptoms, medications, and the date of your last checkup. Ask which preventive care is relevant to your age, health history, and country. Screening can include more than blood tests, such as blood-pressure checks and appropriate cancer screening. MyHealthfinder explains how to discuss screening.

Keep daily prevention in the picture too: food, physical activity, sleep, smoking, alcohol, mental health, and recommended vaccinations. A testing plan should support those conversations, not replace them.

Plan blood testing with a purpose.

Ask a clinician whether blood tests would help establish a useful baseline, investigate a symptom, or monitor a condition or medication. Those are different reasons for testing. There is no single panel or repeat schedule that everyone needs.

• Cholesterol and blood sugar: a lipid profile and appropriate glucose testing may be part of a cardiovascular or diabetes-risk assessment.
• Blood count, kidney, or liver tests: discuss these in relation to your history, symptoms, conditions, and medications, rather than treating them as a compulsory package.
• Thyroid, iron, vitamins, or hormones: ask what specific question a proposed test would answer and whether there is a reason to suspect a problem.

The NHS overview of blood tests explains common tests and why they are used. More testing is not automatically better prevention. For example, the USPSTF finds insufficient evidence to assess routine vitamin D screening in asymptomatic adults; that is different from testing someone with a clinical reason. Read the recommendation and its limits.

Before ordering a large commercial panel, ask what each result could change, how false positives would be handled, and what follow-up might cost. Follow the laboratory's preparation instructions. Do not fast or stop medication just because an AI tells you to.

Help me prepare a prevention review, including whether blood testing would be useful. Read master-health.md and my checked records. Ask for missing context: age, relevant anatomy, country, symptoms, diagnoses, medications, pregnancy where relevant, family history, previous tests, and my priorities.

Separate screening when I have no symptoms from tests to investigate symptoms or monitor a known condition or medication. Do not give everyone the same panel.

For each test or screening worth discussing, explain the question it answers, why it might apply to me, whether I have already had it, possible benefits and false positives, and what decision a result would change. Use current local guidance and relevant studies, and explain uncertainties or disagreements. Do not set a blanket yearly testing schedule.

Consider cardiovascular and diabetes risk assessment where appropriate. Only propose additional blood count, kidney, liver, thyroid, iron, vitamin, hormone, or other tests when there is a clear rationale to discuss. Flag commercial panels with unclear clinical usefulness.

Include prevention beyond blood tests: sleep, movement, food, smoking and alcohol, mental health, blood pressure, vaccinations, and appropriate screening discussions. Do not treat normal blood results as a guarantee of health.

Create topics/prevention.md with questions for my clinician and a section for the plan we actually agree on. Leave dates, tests, and repeat intervals undecided until confirmed. Add this conversation to the master index. Do not order tests, invent a personal plan, or recommend stopping medicines for test preparation.

Make the results useful over time.

Add the original report to your folder, run the record prompt, and check the extraction. Ask for a dated comparison that keeps units, reference ranges, laboratory differences, and relevant circumstances visible. Agree with your clinician which findings need follow-up and when to repeat a test.

A result outside a reference range does not establish a diagnosis on its own, and a result inside it does not rule out every problem. Persistent symptoms still deserve attention. Avoid chasing an AI-generated “optimal” range without evidence that doing so improves health. MedlinePlus explains how to interpret laboratory results.

Save the agreed plan in the Prevention file and bring a short summary into the master document. Record who recommended each test or follow-up and why.

Go deeper into genetic testing.

Blood results show measurements taken at a particular time. Genetic testing asks a different set of questions about inherited variation. Bringing the two together can help you prepare more specific questions about risk, medication response, and family history. It does not mean every genetic difference needs an intervention.

You can explore genetic testing as part of your prevention discussion, especially when your personal or family history raises a question. You do not need whole-genome sequencing to use this guide, and it is not a routine requirement for everyone.

Understand what kind of test you are considering.

Consumer genotyping

Checks selected DNA positions. An ancestry or trait result is not the same as a comprehensive clinical assessment, and an untested variant cannot be ruled out.

Targeted clinical testing

Examines a particular gene, group of genes, or type of variation for a defined question. A focused test may be more appropriate than the broadest available test.

Whole-exome sequencing

Focuses on protein-coding regions. It does not cover all the genetic material outside those regions.

Whole-genome sequencing

Surveys coding and noncoding regions more broadly. Coverage, method, and interpretation still have limits. It does not reliably detect or explain every possible genetic change.

MedlinePlus explains exome and genome sequencing. Broader testing can also find unexpected results or variants whose significance is not known. Ask about those possibilities before choosing a test.

Compare providers on the actual laboratory method and validation, which findings they report, access to genetic counseling, confirmatory testing, data export, total costs, and future reinterpretation. Read their storage, deletion, research-use, and sharing policies. Do not choose only by the number of reports advertised.

Know which files to keep and which to read first.

PDF or other human-readable reports

Start here. Keep the laboratory's interpretation, test method, limitations, and report date. Ask ChatGPT to explain the report before attempting additional analysis.

SNP exports and VCF files

Contain genotype results or variant calls. They need the right reference-genome version and technical context. A VCF is not a list of diagnoses; an absent entry does not necessarily mean a variant was ruled out.

FASTQ files

Contain sequencing reads and quality information. Keep a backup if supplied. Raw-data analysis needs suitable tools and quality checks, not just a conversational summary.

BAM or CRAM files

Store sequence-alignment data. These can support specialist analysis but are not ready-made health reports. A chat saying it read a filename is not evidence that it analyzed the data.

Technical definitions are available in the HTS format specifications. Do not paste huge raw files into a chat or strip a variant list down to alarming labels without understanding the method. Ask a qualified professional which files and validated analysis are appropriate for the question.

Five genetics questions worth exploring carefully.

Create a Genetics conversation and begin with the report you have. Use the prompts that match your questions. If medication response or nutrition becomes a substantial topic, give it its own conversation and cross-reference the saved files.

Inherited risk and unexpected findings. Ask what the laboratory actually found and whether it could matter for your care or family. A risk association is not a diagnosis. A variant of uncertain significance should not become the basis for a preventive procedure or treatment change. The NCI explains result categories in inherited cancer testing.

Read these genetic reports: [filenames], along with my master health document and relevant family history. Tell me exactly what you could read and what kind of test was performed.

For each reported finding, record the gene, variant and genotype only as provided, source page, laboratory interpretation, date, and test limitations. Distinguish pathogenic or likely pathogenic findings, variants of uncertain significance, carrier results, and common risk associations. Do not treat a variant of uncertain significance as a diagnosis or an actionable result.

Where useful, check current clinical sources and ClinVar, including review status, dates, and conflicting submissions. Do not treat a database match as a clinical diagnosis. Explain the difference between a risk factor, a confirmed diagnosis, and an incidental or secondary finding.

Do not infer missing variants or a negative result from the absence of a VCF entry. Do not calculate my lifetime risk without an appropriate validated basis, and explain any population limitations.

Prepare questions for a genetic counselor about clinical confirmation, family implications, and whether a verified finding could change screening or care. Do not prescribe a screening schedule from raw data. Save the dated explanation in topics/genetics.md and update the master index without turning unconfirmed findings into facts.

Pharmacogenomics. Some genetic results can help clinicians choose or dose particular medicines, alongside other clinical information. They do not predict the response to every drug. Keep your actual medication list attached and bring the findings to a prescriber or pharmacist. Read about pharmacogenetic testing; CPIC publishes gene-drug guidelines.

Review my pharmacogenetic report [filename], current medications, supplements, and relevant medical history.

Build a table of the results actually reported: gene, reported genotype or star alleles, laboratory-assigned phenotype, relevant medication, source page, and questions for my prescriber or pharmacist. Mark missing results as unknown. Do not infer a complete drug-metabolism profile from a few SNPs.

Consult current CPIC guidelines or other appropriate clinical sources for the specific gene-drug pairs. Explain what the evidence does and does not establish, other factors such as interactions or kidney and liver function, and whether confirmatory testing is needed.

Do not tell me to start, stop, substitute, or change the dose of a medicine. Do not label a drug safe or unsafe from genotype alone. Save a dated medication-response research file under topics/, linked from the master index, and a short list of questions for a professional medication review.

Carrier status and family planning. A carrier result can raise questions about inheritance and reproductive planning. What it means depends on the condition, inheritance pattern, and partner information when relevant. A negative result leaves some residual risk. Discuss the scope and next steps with a genetics professional rather than assuming a genome file tests for every condition.

Explain my carrier-screening report [filename] and its stated scope. Use only the findings actually reported.

For each result, explain the condition, inheritance pattern, test limitations, and the questions it raises for a genetic counselor. Explain residual risk after a negative result. Do not assume that a whole-genome dataset rules out conditions requiring specialized methods.

If reproductive planning is relevant, explain what additional information or partner testing a genetics professional might discuss. Do not apply a single inheritance percentage to all conditions or assume a partner's genotype.

Keep another person's information separate and use it only with appropriate permission. Save the explanation, sources, date, and unanswered questions under topics/ and add a brief entry to the master index. Do not make reproductive decisions for me.

Nutrigenomics, sleep, and lifestyle. Explore claims about nutrition, caffeine, exercise response, or chronotype, but ask whether changing behavior based on that result has actually improved meaningful outcomes. Your habits, symptoms, and relevant tests remain important. A DNA report should not automatically determine your diet or a precise supplement dose.

Research this genetics-related nutrition or lifestyle claim: [claim or report section].

Read the actual report, my relevant history, habits, and checked laboratory results. Do not invent a genotype. Compare the report's marketing claim with current studies that examined the specific variant, intervention, and outcome.

Explain association versus causation, effect size, population differences, replication, and whether genotype-guided changes have improved meaningful outcomes compared with ordinary personalized advice. Include credible contrary findings and conflicts of interest.

Consider diet, caffeine, sleep, activity, and other relevant lifestyle factors holistically. Do not derive a precise macronutrient split, caffeine cutoff, supplement dose, or exercise prescription from a variant alone.

Separate established findings from preliminary ideas and unsupported personalization. Suggest questions for a qualified clinician or dietitian. Save the dated research and sources in the relevant topic file and update its master-index entry.

Methylation and MTHFR. Keep this topic, but investigate it properly. A report of inherited variants is different from a test measuring DNA methylation. Common MTHFR variants do not mean a person cannot process folic acid; the CDC addresses that claim directly. Ask for evidence before turning a pathway diagram into a diagnosis or a supplement stack.

Help me examine this methylation-related report or claim: [filename or claim].

Explain what was actually measured. Distinguish inherited DNA variants from a test measuring DNA methylation. Discuss MTHFR or other genes only when relevant, and do not infer missing genotypes or a general methylation disorder.

Assess the evidence connecting the finding to my question. For claims involving folate, B12, homocysteine, mood, or fatigue, distinguish a biological mechanism from evidence that a specific intervention benefits people like me. Consult current research, including evidence that challenges the claim.

Check claims that a common MTHFR variant means folic acid cannot be processed against the CDC's guidance. Do not prescribe methylfolate, methylated B vitamins, or a supplement stack from genotype alone. Do not recommend avoiding folic acid based only on a common MTHFR variant.

Explain what, if anything, would justify discussing further tests with a clinician, what remains unknown, and possible harms of unnecessary testing or supplementation. Save the dated sources and findings in the relevant topic file, preserving uncertainty in the master index.

Bring verified findings back into the bigger picture.

Keep the original report and date beside each interpretation. If an important finding comes from consumer raw data or a third-party analysis, ask about clinical confirmation before acting. A negative consumer result does not exclude every risk. The FDA explains consumer-test limitations.

Add questions to the relevant Sleep, Food, Medication, or Prevention conversation. Update the master only with a clear distinction between a confirmed finding, a hypothesis, and a recommendation your clinician has actually made.

Genetic data also concerns your relatives and may remain identifying without your name. Consider unexpected findings, family implications, and privacy before testing or sharing. Review the risks and limitations of genetic testing.

Check the answer before acting on it.

If an answer worries you, ask it to show the exact report and page behind the claim. Check the original value, units, date, and comments. A correct extraction and a correct interpretation are two different things.

If it made a mistake, correct the saved record and any topic or master summary that used it. Another AI agreeing with the answer is not proof. Bring important findings and the original report to a qualified clinician.

Do not start, stop, or change medication based on an AI answer. Use this guide for understanding and preparing questions. It is not an emergency service; seek urgent medical help when needed rather than waiting for research.

Keep control of what you share.

Before adding medical records, review the data controls and terms for your own AI account, including model training, retention, and deletion. Paying for a personal plan does not automatically give you the same protections as a business service.

Your folder is on your computer, but relevant excerpts, prompts, and tool results may be sent to OpenAI when ChatGPT works with them. This is not an offline-only setup. OpenAI explains local processing and its boundaries.

Share only what is needed. If you remove identifying details from a copy of a report, use a real redaction tool and check hidden text and metadata. Covering text with a black rectangle may leave it recoverable. Keep the original unchanged. Read Adobe's redaction instructions.

Protect your computer and AI accounts, and check who can access any backup or cloud-synced folder. Deleting a local file does not necessarily remove copies already in chats or backups. For professional use, follow your organization's approved process for medical records.

Decode does not ask you to upload health information here. This website loads Google Analytics, Google Fonts, and styling from a Tailwind CDN; those services receive connection and browser information. Your health conversations happen in the tool you choose.