--- title: "Teach AI your content with context files instead of fine-tuning a model" description: "Skip model training and load your content as context with a CLAUDE.md file, a vault of past work, and an assistant that reads the files." canonical: "https://scalewithsearch.com/articles/how-to-train-ai-on-my-content" date: "2026-01-28" modified: "2026-09-25" --- ## Site navigation - [Scale With Search](https://scalewithsearch.com/) - Real estate - Real estate - [Real estate](https://scalewithsearch.com/for/real-estate) - Work - Start here - [Send your brief](https://scalewithsearch.com/work#send-your-brief) - [Prepare your six-question brief](https://scalewithsearch.com/work#prepare-your-six-question-brief) - Build - [Site build, content library with SEO, signal desk](https://scalewithsearch.com/work) - For your business - Trades and home services - [Auto body and collision shops](https://scalewithsearch.com/for/auto-body-and-collision-shops) - [Foundation and home repair contractors](https://scalewithsearch.com/for/foundation-and-home-repair) - [Garage door and fencing contractors](https://scalewithsearch.com/for/garage-door-and-fencing-contractors) - [HVAC contractors](https://scalewithsearch.com/for/hvac-contractors) - [Janitorial and commercial cleaning companies](https://scalewithsearch.com/for/janitorial-and-commercial-cleaning) - [Locksmiths](https://scalewithsearch.com/for/locksmiths) - [Moving companies](https://scalewithsearch.com/for/moving-companies) - [Pest control companies](https://scalewithsearch.com/for/pest-control-companies) - [Plumbing and electrical contractors](https://scalewithsearch.com/for/plumbing-and-electrical-contractors) - [Restoration and water or fire damage companies](https://scalewithsearch.com/for/restoration-and-water-fire-damage) - [Roofing companies](https://scalewithsearch.com/for/roofing-companies) - [Towing companies](https://scalewithsearch.com/for/towing-companies) - [Tree services and landscaping companies](https://scalewithsearch.com/for/tree-services-and-landscaping) - [Solar installers](https://scalewithsearch.com/for/solar-installation) - [General contractors](https://scalewithsearch.com/for/general-contractors-and-construction) - [Paving, concrete, and flooring contractors](https://scalewithsearch.com/for/paving) - Practices and professional services - [Bookkeeping and tax practices](https://scalewithsearch.com/for/bookkeeping-and-tax-practices) - [Dental practices](https://scalewithsearch.com/for/dental-practices) - [Family and criminal defense law firms](https://scalewithsearch.com/for/family-and-criminal-defense-law-firms) - [Med spas and aesthetics practices](https://scalewithsearch.com/for/med-spas-and-aesthetics) - [Personal injury law firms](https://scalewithsearch.com/for/personal-injury-law-firms) - [Veterinary clinics](https://scalewithsearch.com/for/veterinary-clinics) - [Gyms and fitness studios](https://scalewithsearch.com/for/fitness) - [Therapy and outpatient health practices](https://scalewithsearch.com/for/therapy-and-outpatient-health) - [Medical billing companies](https://scalewithsearch.com/for/medical-billing) - [Insurance agencies](https://scalewithsearch.com/for/insurance-agencies) - [Financial advisors](https://scalewithsearch.com/for/financial-advisors) - [Property management companies](https://scalewithsearch.com/for/property-management) - [Recruiting and staffing agencies](https://scalewithsearch.com/for/recruiting-and-staffing) - [Architects and interior designers](https://scalewithsearch.com/for/architects-and-interior-designers) - [Logistics and supply chain companies](https://scalewithsearch.com/for/logistics-and-supply-chain) - Agencies, MSPs, and manufacturing - [IT and managed service providers](https://scalewithsearch.com/for/it-and-managed-service-providers) - [Machine shops and precision manufacturers](https://scalewithsearch.com/for/machine-shops-and-precision-manufacturing) - [Marketing agencies and freelancers](https://scalewithsearch.com/for/marketing-agencies-and-freelancers) - [SEO agencies and consultants](https://scalewithsearch.com/for/seo-agencies-and-consultants) - [Small manufacturers and fabricators](https://scalewithsearch.com/for/small-manufacturers-and-fabricators) - Restaurants, shops, studios, and nonprofits - [Restaurants and hospitality businesses](https://scalewithsearch.com/for/restaurants-and-hospitality) - [Retail stores and ecommerce sellers](https://scalewithsearch.com/for/retail-and-ecommerce) - [Photographers, event planners, and travel agents](https://scalewithsearch.com/for/photographers) - [Churches and nonprofits](https://scalewithsearch.com/for/churches-and-nonprofits) - [All industries](https://scalewithsearch.com/for/) - Learn - For your office - [Office job guides](https://scalewithsearch.com/guides/) - [Browser calculators](https://scalewithsearch.com/tools/) - Start here - [How it works](https://scalewithsearch.com/how-it-works) - [Free Starter Kit](https://scalewithsearch.com/kit/business-memory-starter-kit.zip) - [Synthetic specimen](https://scalewithsearch.com/specimen/working-session-specimen.zip) - Guides - [The Complete Guide to Business Memory for AI Agents](https://scalewithsearch.com/articles/business-memory-for-ai-agents-guide) - [The Complete Small-Business Guide to AI Agent Governance](https://scalewithsearch.com/articles/ai-agent-governance-guide-small-business) - [The Complete Guide to Leaving Vendor AI Memory](https://scalewithsearch.com/articles/leaving-vendor-ai-memory-guide) - Articles by cluster - [Business memory](https://scalewithsearch.com/articles/business-memory-for-ai-agents-guide) - [Agent governance](https://scalewithsearch.com/articles/ai-agent-governance-guide-small-business) - [Migration and ownership](https://scalewithsearch.com/articles/leaving-vendor-ai-memory-guide) - For machines - [llms.txt](https://scalewithsearch.com/llms.txt) - [llms-full.txt](https://scalewithsearch.com/llms-full.txt) - [Machine view](https://scalewithsearch.com/?view=machine) - Company - Evidence - [Proof](https://scalewithsearch.com/proof) - Company - [About](https://scalewithsearch.com/about) # Teach AI your content with context files instead of fine-tuning a model. You have years of proposals, audits, articles, and email. You want an AI assistant to write the way you write and follow the structure you use. So you ask, "How do I train AI on my content?" For most businesses, that is the wrong question. You do not need to train a model. You need to give it context. Training sounds like fine-tuning, custom models, and machine learning infrastructure: expensive, technical, and slow. Context loading is a Markdown file and a folder. You likely have the tools already. ## Know the difference between training and context loading Training a model means you take a base model, such as a GPT or Claude model, and adjust its internal weights with your data. You prepare thousands of examples, run a training job, and get a custom version of the model that has absorbed your content. | Factor | Fine-tuning | Context loading | |---|---|---| | What changes | The model's internal weights | The input the model reads in each session | | Cost | Far more than context loading: data preparation, training runs, and hosting | No cost beyond your AI plan | | Time | Weeks to prepare data, train, test, and iterate | One setup session | | Skills | API access and machine learning experience | A text editor and a folder | | Updates | Frozen at training time; new content needs retraining | Current as soon as you save a file | Context loading gives the assistant your content as input during the session. Nothing is trained. The assistant reads your files every session, so it always sees the current version. ## Know when fine-tuning fits Fine-tuning suits a product that runs the same narrow task for thousands of users, where loading context on every request costs too much. A company that runs one task millions of times can justify a trained model. A consultant, coach, agency owner, or freelancer does not need a custom model. You need an assistant that knows your business, your clients, and your content. Context loading gives you that without the cost, complexity, or maintenance of fine-tuning. ## Understand how context loading works A model has a context window: the amount of text it can process in one session. In January 2026, Claude Opus 4.5 had a 200,000-token window. That is roughly 150,000 words, or about 500 pages. By September 2026, the current Claude models listed a window of 1 million tokens. Check the current figure for the model you use; the [context window comparison](/articles/context-window-vs-persistent-ai-memory) explains what the window holds and what it does not. With Claude Code and an Obsidian vault, Claude can open any file in the vault and read it into the context window. A typical vault holds: - `CLAUDE.md`, with your business details, voice rules, and constraints; - client briefs, with past projects, deliverables, and outcomes; - a content library of posts, articles, and email you wrote; - templates for proposals, contracts, and outreach; - meeting notes from calls, strategy sessions, and decisions. Claude does not memorize these files the way a trained model absorbs data. It processes them as input, the same way it processes your prompt. The practical effect is similar: it works from your content. ## See it in practice Suppose you run a consultancy that designs onboarding flows for SaaS companies. You have written more than 50 onboarding audits over two years. They live in your vault at `/Client Work/Onboarding Audits/`. You ask Claude: "Write an onboarding audit for a new client that sells project management software to construction companies." Claude reads `CLAUDE.md`, which explains your audit structure. It reads five past audits for similar clients. It reads your template file with the standard sections and questions. The draft follows your format, uses your language, asks your questions, and keeps your structure. You did not train a model. You gave it the correct files. Past client work needs a boundary. Remove client names and confidential details from examples, or keep each client's files where only that client's work can read them. The [client context guide](/articles/separate-client-context-before-agent-work) shows how to separate them before an agent starts. ## Set it up in five steps 1. Install Obsidian and create a vault. Obsidian is a Markdown note app, and a vault is a folder on your computer. The [vault setup guide](/articles/how-to-set-up-ai-vault) covers this step in detail. 2. Install Claude Code, Anthropic's command-line tool for Claude. In January 2026, it required a paid Claude plan, and one documented install command was `npm install -g @anthropic-ai/claude-code`. Check Anthropic's current setup page before you install. Start it with `claude` from inside the vault folder. 3. Create `CLAUDE.md` at the vault root. Write in plain language. Include: - who you are: name, business, role, industry, and clients; - what you do: services, deliverables, and timelines, with prices kept in a separate rate file; - voice: tone, style, banned phrases, and samples of your writing; - constraints: what you do not do, whom you do not work with, and deal structures you refuse; - examples: two or three samples of your best work. 4. Add your content to the vault: posts, client deliverables, email templates, strategy documents, and case studies. Organize by client, project type, or date. Obsidian indexes all of it. 5. Use it. Open Claude Code and make a request. Claude reads `CLAUDE.md` automatically, and it reads any file you name. Example: "Read the onboarding audit I did for [client] and create a similar one for this new client." ## Know what you can load Any text in Markdown can serve as context: | Type | Examples | |---|---| | Writing samples | Posts, articles, social threads, newsletters | | Client work | Proposals, audits, strategy decks, reports, presentations converted to Markdown | | Templates | Email scripts, outreach sequences, contract language, SOPs | | Research | Industry notes, competitor analysis, market research, customer interviews | | Personal knowledge | Book notes, frameworks, mental models, decision logs | Brand voice needs extra care. Old samples can carry old claims and old positioning. The [brand memory guide](/articles/ai-brand-memory-content-creation) explains how to keep approved voice without freezing outdated claims. ## Estimate how much fits With a 200,000-token window, about 150,000 words fit in one session. A `CLAUDE.md` file may be 1,000 words. A typical post is 1,500 words. A client audit is 3,000 words. You can load 30 to 50 documents in one session if you need to, and most tasks need far fewer. Claude Code can search the vault and pick files. You can ask, "Read all my onboarding audits and summarize the common mistakes I find." Claude finds the files, reads them, and writes the summary. ## Compare context loading with RAG Retrieval-augmented generation (RAG) searches a database for relevant chunks of text, retrieves them, and uses them as context. It suits very large collections of thousands of documents. It needs vector databases, embedding models, and a search layer. Context loading is simpler. Claude reads the files directly, with no database, embeddings, or search layer. If the documents that matter number 50 to 200, not 10,000, Claude can read them directly. When the collection outgrows direct reading, [RAG is retrieval, and business memory also needs authority](/articles/rag-vs-business-memory) explains what to add. ## Keep the content live A fine-tuned model learns from a snapshot. If you fine-tune in January, the model does not know what you write in February. You retrain to update it. A vault is live. Add a file, and Claude can read it in the next request. You do not retrain, re-upload, or version a model. ## Test the output against your samples Check that context loading works before you rely on it: 1. Pick three recent pieces you wrote and liked. 2. Ask Claude for three new pieces of the same type, with only the vault as context. 3. Compare structure: sections, order, and length. 4. Compare language: banned phrases, sentence length, and terms. 5. Record each gap and add the missing rule or sample to `CLAUDE.md`. 6. Repeat in a new session. The test passes when the new drafts follow your structure and rules without a correction from you. ## Scale to a larger library Three years into a business, a vault can hold a large library. One example: 120 client deliverables, 80 posts, 40 email templates, 30 case studies, 15 SOPs, and more than 200 meeting notes. That is about 500 documents and thousands of pages. Claude reads what each request needs, not all 500 at once. You ask for a proposal for a new fintech client. Claude reads `CLAUDE.md`, three past fintech proposals, your proposal template, and your rate file. The draft matches your structure, uses your language, includes the correct prices from the rate file, and cites relevant case studies. You did not train a model on 500 documents. You organized your files and told Claude where to look. ## Start with ten examples Install Obsidian and create a vault. Write `CLAUDE.md` with your business details, voice rules, and examples. Install Claude Code and start it in the vault folder. Copy 5 to 10 examples of your best work into the vault: client deliverables, email, proposals, or whatever represents your output. Ask Claude to create something similar, and watch which files it reads. The drafts show you what the vault already teaches and what it still lacks. ## Related: AI memory - [Write an email context file so AI drafts sound like you](/articles/ai-memory-for-email-writing) - [Write a social media context file so AI posts sound like your account](/articles/ai-memory-for-social-media) - [Give AI your business context so it stops writing generic answers](/articles/why-ai-gives-generic-answers) ---- ```text .|########||. .|########||. .|########||. |##||. .||##|. |##||. .||##|. |##||. .||##|. |#|. .|#|. |#|. .|#|. |#|. .|#|. |#| |#| |#| |#| |#| |#| .#| |#. .#| |#. .#| |#. .#. .#| .#. .#| .#. .#| |#. .#| |#. .#| |#. .#| |# ...... #| |# ...... #| |# ...... #| .# ||#########| #| .# ||#########| #| .# ||#########| #| .#. |######||######|. .#. .#. |######||######|. .#. .#. |######||######|. .#. #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# # .||####|###########################|. # # .||####|###########################|. # # .||####|###########################|. # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || #. .######|#|#########| .# #. .######|#|#########| .# #. .######|#|#########| .# .#. |#####||||#####| .#. .#. |#####||||#####| .#. .#. |#####||||#####| .#. |# ||########|| #. |# ||########|| #. |# ||########|| #. |# ...... #| |# ...... #| |# ...... #| |#. .#| |#. .#| |#. .#| |#. .#. |#. .#. |#. .#. .#| |#. .#| |#. .#| |#. |#| |#| |#| |#| |#| |#| .|#|. .|#|. .|#|. .|#|. .|#|. .|#|. |##||. .||##| |##||. .||##| |##||. .||##| .||########||. .||########||. .||########||. Scale With Search 2026 [scalewithsearch.com](https://scalewithsearch.com) ```