--- title: "Move business context out of custom instructions and into files" description: "Custom instructions are too small, too static, and too universal to hold a business. Learn what belongs in them and what belongs in context files." canonical: "https://scalewithsearch.com/articles/why-custom-instructions-fail" 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) # Move business context out of custom instructions and into files. When ChatGPT launched custom instructions in 2023, each field held 1,500 characters. That is about 250 words, or three short paragraphs. Limits can differ by plan and change over time, so check your current settings. Many owners try to fit the whole business into that box: role, clients, services, communication style, pricing structure, and workflows. It does not fit, and the parts that do fit go stale. Custom instructions have five structural problems for business context. This page explains each one, then shows what to keep in the field and what to move into files. ## See the space problem Do the arithmetic for a freelance consultant. The AI needs to know five things: | Context | Approximate length | |---|---| | Services offered | 150 words | | How work is priced | 100 words | | Client qualification criteria | 100 words | | Communication style | 75 words | | Standard email templates and responses | 200 words | | **Total** | **625 words** | That is more than double the 250 words of the original limit. So the consultant cuts. The pricing details go. The qualification criteria go. Only the most critical lines stay. A higher limit delays the cut, but client lists and project history grow every month. Now the AI knows that the user is a consultant, but not how the consultant works. It cannot price a project, qualify a lead, or draft an email in the consultant's voice, because the examples did not fit. The cuts removed the context that made the output useful. ## See the static problem Custom instructions do not update themselves. You sign a new client, and the instructions still list the old one. You change your pricing, and the instructions still show last month's model. You refine your services, and the instructions describe what you used to do. Each change needs a manual edit of the text field. The field is one block of text with no structure, so an edit often means a rewrite. Most people set custom instructions once and leave them. Six months later, the AI works from facts that are no longer true. ## See the universal problem Custom instructions apply to every conversation in the same way. A client proposal, a code fix, a personal email, and a recipe question all get the same context. That creates two failure modes: - **Too specific.** Instructions written for business work make every chat assume business work. A request for help with a family trip turns into something shaped like a client deliverable. - **Too generic.** Instructions vague enough to fit everything give no task enough context to help. Real work needs context by domain. Client work needs client facts. Code work needs your tech stack. Personal tasks need personal preferences. One instruction set cannot serve all of them. ChatGPT Projects now add instructions per project, which narrows the problem. The project instructions still live inside one product and one project. The [comparison of custom and project instruction limits](/articles/chatgpt-custom-vs-project-instructions-limits) covers what each layer holds. ## See the long-chat problem A common belief says that a long conversation drops custom instructions first. That is not how it works. The product keeps the instructions in the context of each conversation. The real effect is weaker. As a chat grows, the model can weight early context less and recent messages more. Research on long contexts found that models use information at the start and end of a long input more reliably than information in the middle (Liu et al., 2023, "[Lost in the Middle](https://arxiv.org/abs/2307.03172)"). In practice, a style rule that held for ten messages can slip by message forty. The model starts to give generic advice that ignores the setup. Instructions are also guidance, not enforcement. A sentence such as "never send an email without my approval" does not block a send tool. The [guide on custom instructions as controls](/articles/chatgpt-custom-instructions-not-hard-constraints) shows how to build a check that does. ## See the composition problem You cannot layer custom instructions. Suppose the AI needs four kinds of context: - who you are (identity) - what business you run (domain) - how you communicate (style) - what you are working on now (state) With custom instructions, all four go into one field. You cannot turn a piece on or off. You cannot update one section without the risk of damage to the others. You cannot keep different sets for different kinds of work. Context files separate the four: | File | Contents | When it loads | |---|---|---| | `IDENTITY.md` | Who you are | Every conversation | | `BUSINESS.md` | What you do | Work tasks | | `STYLE.md` | How you communicate | Writing tasks | | `PROJECTS.md` | Current work | When relevant | Each file has its own owner, update date, and scope. You load the files a task needs and leave the rest out. ## Recognize the two common patterns Most people end up in one of two places. The first group writes a short line: "I'm a marketing consultant. I work with small businesses. Be concise and professional." Anything longer does not fit. They then re-explain the client, the product, and the brand voice at the start of every conversation. The custom instructions did almost nothing. The second group fills the field to the limit, hits it, and deletes important lines to make room for new ones. The field becomes a record of which facts lost the last edit. Both groups work against the tool instead of with it. ## Know what custom instructions were built for Custom instructions were never a memory system. They are a convenience for basic preferences, so you do not type "be concise" in every chat. Vendors know the gap. OpenAI added a separate memory feature. Anthropic added Projects that hold documents. Those features help, but they stay inside one platform, and the vendor controls how they store and select your context. A business needs context it owns. That means files that you write, organize, and load when a task needs them. No product team sets their character limit. ## Sort each line: preference, fact, or rule Run this audit on your current custom instructions. 1. Copy every line of your custom instructions into a document. 2. Mark each line as a preference, a fact, or a rule. 3. Keep the preferences in custom instructions. Examples: tone, output format, length, language. 4. Move each fact into the context file for its domain. Examples: clients, services, pricing model, projects. 5. Move each rule into a file, and add a check where the rule matters. Examples: approval steps, banned claims. 6. Open a new chat and run one real task with the files attached or loaded. 7. Compare the output with the output from the old setup. The preferences that stay usually fit in a few sentences. An illustrative case shows the result. A business consultant has 250 words in custom instructions. Half of them are client names and project details that change every week. He moves that half into `CLIENTS.md`. The file grows to about 2,000 words, organized by client, with project status and notes. He updates a client entry in seconds. The file loads when he works on client tasks. His custom instructions shrink to "I'm a business consultant. Be direct. Use examples." ## Compare the two storage places | Property | Custom instructions | Context files | |---|---|---| | Size | Fixed character limit per field | No fixed limit; each loaded line uses context, so keep files focused | | Updates | Edit one block of text in settings | Change one line in the right file and save | | Scope | Every conversation | Only the tasks that load the file | | Composition | One undivided field | Several files, loaded together as needed | | Portability | Inside one product | Plain text that any AI tool can read | | Long chats | Can lose weight as the chat grows | Same risk inside one chat; reloads at full strength in each new session | Files do not beat the model's attention limits. A file loaded at the start of a long chat can lose weight too. The difference is that each new session loads the current file in full, so a restart restores the context. For a file that loads by itself at each session start, see the [CLAUDE.md template for business context](/articles/claude-md-template-business-context). ## Keep the right line in each place Use custom instructions for what they do well: tone, output format, and general style. Put everything else in context files: business details, clients, processes, and current projects. The real context then lives where you can maintain it, compare its versions, and move it to another tool. The [persistent context guide](/articles/persistent-context-across-ai-tools) shows how to test that the same files work across products. ## Related: AI memory - [Stop AI from inventing client details with a verified client file and a stop rule](/articles/ai-keeps-hallucinating-my-details) - [Find where Claude Code /init stops and plan the memory it leaves out](/articles/claude-code-init-limitations) - [Copy a CLAUDE.md template for your role and adapt it in 30 minutes](/articles/claude-md-template-examples) ---- ```text .|########||. .|########||. .|########||. |##||. .||##|. |##||. .||##|. |##||. .||##|. |#|. .|#|. |#|. .|#|. |#|. .|#|. |#| |#| |#| |#| |#| |#| .#| |#. .#| |#. .#| |#. .#. .#| .#. .#| .#. .#| |#. .#| |#. .#| |#. .#| |# ...... #| |# ...... #| |# ...... #| .# ||#########| #| .# ||#########| #| .# ||#########| #| .#. |######||######|. .#. .#. |######||######|. .#. .#. |######||######|. .#. #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# # .||####|###########################|. # # .||####|###########################|. # # .||####|###########################|. # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || #. .######|#|#########| .# #. .######|#|#########| .# #. .######|#|#########| .# .#. |#####||||#####| .#. .#. |#####||||#####| .#. .#. |#####||||#####| .#. |# ||########|| #. |# ||########|| #. |# ||########|| #. |# ...... #| |# ...... #| |# ...... #| |#. .#| |#. .#| |#. .#| |#. .#. |#. .#. |#. .#. .#| |#. .#| |#. .#| |#. |#| |#| |#| |#| |#| |#| .|#|. .|#|. .|#|. .|#|. .|#|. .|#|. |##||. .||##| |##||. .||##| |##||. .||##| .||########||. .||########||. .||########||. Scale With Search 2026 [scalewithsearch.com](https://scalewithsearch.com) ```