--- title: "Stop AI from inventing client details with a verified client file and a stop rule" description: "Stop AI from inventing client names and project details with a verified client file, a rule to stop on unknowns, and a fact check per draft." canonical: "https://scalewithsearch.com/articles/ai-keeps-hallucinating-my-details" 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) # Stop AI from inventing client details with a verified client file and a stop rule. You ask Claude to draft an email to a client. It returns a message addressed to "Sarah Johnson" at "Acme Industries." You have no client named Sarah. You have never worked with Acme Industries. The model made both up. This happens because a large language model completes patterns. When your request leaves gaps, the model fills them with plausible text drawn from its training data. It does not know your clients, your projects, or how your business is structured, so it invents them. ## Why the model fabricates business details The root cause is missing data. By default, each conversation starts without your business. You may have described your three main clients last week, but today's session has no record of that exchange. Ask for "an update for the Johnson account," and the model sees a name it cannot resolve. Many models do not stop and admit the gap. They generate the most common patterns for "Johnson" and "account." The result sounds professional and contains no accurate information about your Johnson account. The problem compounds with repetition. You correct the model, it apologizes, and it writes another draft. That draft fixes the name and invents project details. You correct it again. The cycle continues, because nothing records your corrections for the next session. ## What fabricated details cost Each fabrication takes correction time. You spot the wrong name and fix it. You find invented project dates, so you review every date. You find services you do not offer, so you verify the proposal line by line. The time adds up. In one example, five minutes of correction on each of ten AI-assisted tasks a day is fifty minutes a day. That is over four hours a week and about 200 hours over a working year. The larger cost is risk. Miss one invented detail in a client email, and you send false information under your name. Let one invented deadline into a proposal, and you have committed to a timeline you cannot meet. The model's confident tone makes these errors easy to miss in a quick review. ## What does not work Most people try the same fixes. They write long prompts that explain the business. They paste client lists into the chat. They keep a "system prompt" in a notes app and paste it into each conversation. These fixes do not last. Text in a chat message ends with the session. Pasted context moves the cost: you spend the time to feed facts to the model instead of to correct its output. Custom GPTs and product memory features help a little. They keep fragments: your company name, maybe a client or two. They lack the structured detail that accurate work needs. They also stay inside one product, so your facts do not follow you from Claude to ChatGPT to another tool. ## Give the model a verified client file A context file is a Markdown document with your real business data: client names, project details, services, and communication preferences. It gives the model facts to use instead of patterns to guess from. The file lives in a fixed location. At the start of each session, the tool loads it before your request. In Claude Code, the root file is `CLAUDE.md`, and it can point to one file per client. Build each client file with the same sections: ```markdown # Client: Martinez Landscaping (fictional) status:: active last_verified:: 2026-09-20 primary_contact:: Elena Martinez, owner preferred_channel:: email, short updates ## Active projects - Spring site redesign: in review; launch planned 2026-10-15 - Monthly newsletter: next issue due 2026-10-01 ## Services we provide - Website maintenance - Newsletter production ## Services we do not provide - Paid advertising - Social media management ## Terms to use - "Site refresh", not "rebrand" ``` When you name a client, the model matches the name against the file and uses the recorded facts. When you ask for a service description, it uses your exact wording. Keep one file per client. A single file with every client invites a second kind of error: a real fact attached to the wrong client. That error is harder to spot than an invented name, because the fact is true somewhere. The guide to [separating client context before an agent touches the work](/articles/separate-client-context-before-agent-work) shows how to load only the client the task names. The guide on [what client data belongs in AI agent memory](/articles/client-data-in-ai-agent-memory) sets limits for personal and sensitive fields. ## Add a rule for unknowns A file reduces fabrication. It does not remove it. If you ask about a client that is not in the file, the model can still invent one. Close that gap with an explicit rule in the root file: ```markdown ## Unknown facts - Use only client names, contacts, dates, prices, and services found in the client files. - If a request names a client, project, or figure that is not in the files, write [UNKNOWN: ] in its place and ask me. - Never invent a name, date, or amount to complete a draft. ``` A marked gap is better than a confident guess. You can fill the gap in seconds. You may not catch the guess at all. Test the rule once. Ask for an update to a client you know is not in the files. Confirm that the draft marks the gap. The list of [stopping rules for business AI agents](/articles/ai-agent-stopping-rules-examples) gives more rules of this kind. ## Check each draft against the file Keep a short check before any client-facing draft leaves the business: 1. List every proper noun in the draft: people, companies, projects. 2. List every date, number, and service named. 3. Find each item in the client file. 4. Mark any item that is not in the file. 5. Fix or remove every marked item before you send. Most of this takes a minute. It catches the error the file could not prevent. If the same kind of error comes back, change the file or the rule, not only the draft. The workflow in [how corrections become part of the system](/articles/corrections-become-system-memory) records the fix so the next run applies it. ## Keep the file verified A client file is only as good as its `last_verified` date. A contact changes roles, a project closes, or you stop a service. Each change turns a verified file into a source of confident, wrong drafts. Set two rules. First, update a client file the day a fact changes, and move the `last_verified` date. Second, at the start of each month, list every file whose date is older than 60 days. Confirm each one against your CRM or your last message thread with that client. Ask the assistant for that list; it can read the dates. Then add a rule to the root file. The rule says: if a client file's `last_verified` date is older than 90 days, say so at the top of any draft that uses it. A draft that opens with "this client file was last verified in June" is a small nuisance. A draft that names a contact who left in June is a sent mistake. ## What this looks like in practice On Monday morning you open Claude Code. The root file loads and points to your client files. You type: "Draft an update email for the Martinez account." The model finds Martinez Landscaping in the client file and reads the active projects. The update names the correct contact, the redesign status, and the next newsletter date. It uses "site refresh," the term you recorded. You then ask for an update to "the Delgado account." There is no Delgado file. The draft says `[UNKNOWN: Delgado account]` and asks which client you mean. No invented name reaches the page. The same file works in other tools. You can paste it into Claude or ChatGPT project instructions, or load it wherever a tool accepts context at session start. You maintain one source of client facts, and each assistant reads from it. Proposals now reference real services. Meeting summaries use correct names. Content drafts reflect the work you actually do. The remaining risk is a stale file, so update the `last_verified` date each time you confirm a client's details. ## Related: AI memory - [Load one context file so AI output stays consistent across sessions](/articles/ai-consistency-problem) - [Structure AI context files so they stay short, current, and readable](/articles/ai-context-file-best-practices) - [Route prompts to the right context file with a keyword table](/articles/ai-context-routing-explained) ---- ```text .|########||. .|########||. .|########||. |##||. .||##|. |##||. .||##|. |##||. .||##|. |#|. .|#|. |#|. .|#|. |#|. .|#|. |#| |#| |#| |#| |#| |#| .#| |#. .#| |#. .#| |#. .#. .#| .#. .#| .#. .#| |#. .#| |#. .#| |#. .#| |# ...... #| |# ...... #| |# ...... #| .# ||#########| #| .# ||#########| #| .# ||#########| #| .#. |######||######|. .#. .#. |######||######|. .#. .#. |######||######|. .#. #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# # .||####|###########################|. # # .||####|###########################|. # # .||####|###########################|. # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || #. .######|#|#########| .# #. .######|#|#########| .# #. .######|#|#########| .# .#. |#####||||#####| .#. .#. |#####||||#####| .#. .#. |#####||||#####| .#. |# ||########|| #. |# ||########|| #. |# ||########|| #. |# ...... #| |# ...... #| |# ...... #| |#. .#| |#. .#| |#. .#| |#. .#. |#. .#. |#. .#. .#| |#. .#| |#. .#| |#. |#| |#| |#| |#| |#| |#| .|#|. .|#|. .|#|. .|#|. .|#|. .|#|. |##||. .||##| |##||. .||##| |##||. .||##| .||########||. .||########||. .||########||. Scale With Search 2026 [scalewithsearch.com](https://scalewithsearch.com) ```