AI Assistants in the Automotive Industry: Technical Documentation and Fault Diagnosis
Compare ChatGPT, Claude, and Gemini with an afternoon test plan after checking their differing vendor-documented file, context, and price limits.
Sources checked 2 Oct 2026
The vendors’ own OpenAI upload FAQ, Claude upload guide, and Gemini file guide settle some file-input details, but they do not establish that ChatGPT, Claude, or Gemini is correct for automotive fault diagnosis or suitable as the final repair authority. Chat Picker has not tested the assistants for this use; this page gives you an afternoon test plan and records only vendor-documented constraints. The earlier version’s unsourced statistics and test results have been removed.
Know these limits before you start
The vendor accuracy pages were read on October 1, 2026. OpenAI says ChatGPT may sound confidently wrong and recommends treating it as a first draft whose quotations, data, technical details, and references are checked. Anthropic says not to use Claude as the sole source, scrutinize high-stakes advice, and inspect original sources. Google says Gemini is probabilistic, may violate its guidelines, and can produce inaccuracies that create physical-safety risks. Keep a qualified human reviewer in the loop.
File limits can determine whether your test resembles a real manual-review task:
- As read on October 1, 2026, OpenAI’s file-upload FAQ sets a 512 MB per-file ceiling, a 2M-token cap for each text or document, and three uploads per day for Free users.
- As read on October 1, 2026, Claude’s upload guide lists a 500 MB chat-file limit but a 30 MB project-file limit. For PDFs from 101 through 1,000 pages, it says Claude processes text rather than visual elements.
- As read on October 1, 2026, Gemini’s file guide allows up to 10 files per prompt and 100 MB per nonvideo file, subject to availability.
On the pricing pages read on October 1, 2026, OpenAI offers a training opt-out on Free, Go, Plus, and Pro; Anthropic offers one on Free, Pro, and Max and says Claude Team is not trained on by default. The Gemini plans page does not state a training setting, so verify the account before using proprietary files. Anthropic’s usage policy says elevated-risk uses should integrate relevant expertise and requires consumer-facing chatbot sessions to disclose that AI, rather than a human, is involved.
As read on October 1, 2026, ChatGPT Plus costs $20 per month, Claude Pro costs $20 monthly or $17 monthly on annual billing with $200 upfront, and Google AI Pro costs $19.99 per month. Confirm the current price before subscribing.
The test plan
Use the same source packet and account settings for each run. Keep one log row per response.
-
Generate a structured service manual. Prepare an authorized extract with headings, a parts table, a torque value, a warning, and page numbers. Prompt: “Create a structured service-manual draft from the attached extract. Preserve headings, part numbers, torque values, warnings, units, and step order. Cite the source page for every statement, write Not in source when unsupported, and add no repair advice.” Check structure, warning placement, unchanged values, and traceability. Log missing content, unsupported additions, and reviewer edits separately.
-
Interpret a fault code. Prepare a separate extract containing one fault code’s definition, prerequisites, test procedure, safety notice, and repair recommendation. Prompt: “Using only the attached extract, separate the fault-code definition, prerequisites, confirmatory checks, possible causes, repair steps, and safety cautions. Cite a page for each and label every inference.” Look for a code definition turned into a diagnosis, unsupported causes, and missing cautions. Log each correction, unsupported claim, and source page.
-
Test a workshop export. Prepare a service bulletin and a blank export template with required headings and a naming convention. Prompt: “Convert the attached bulletin into the attached template without adding technical content. Preserve identifiers, dates, warnings, and units, and mark any missing field.” Check whether the structure survives export and every field remains traceable. Log formatting changes, missing fields, manual rework, and unsupported additions.
-
Calculate cost and ownership. Prepare a worksheet containing the checkout price, billing cycle, expected documents, source pages, test runs, reviewer time, and likely rework. Prompt: “Using only the attached worksheet, calculate cost per completed document and total ownership cost for the stated period. Show each formula, keep units consistent, flag missing inputs, and make no estimate.” Check for mixed billing periods or ignored limits. Log subscription cost, usage, review time, rework, and administrative work using your measured values.
-
Probe limitations. Prepare questions the source can answer and others it cannot. Prompt: “Answer each question using only the attached source. Give a page citation for every answer and write Not in source when the source does not answer it. Do not use web search.” Look for invented citations, confident unsupported answers, altered values, and omitted warnings. Log each error by type, whether review caught it, and the correction required.
How to read your results
Turn the log into a requirement checklist, not a universal score. A run fails when a safety-critical value, warning, or repair step cannot be traced to the supplied source, or when review catches an invented citation. Compare only passing runs using your own records for review effort, workflow fit, usage, data controls, and cost.
Do not treat citations as proof. OpenAI says search results may be incomplete, outdated, or incorrect; Anthropic advises cross-referencing cited sources; and Gemini says its Sources button appears when sources are available. If no run passes, do not use that configuration for repair decisions. Your error counts describe this source packet, not an assistant’s performance in every task.
Where the plans differ
Use the three-way comparison matrix and focus on these rows: Free plan, Low-cost tier, Main paid plan, Heavy-use plans, Team plan, Context window in the app, How usage limits are described, Training on your chats, and Ads.
Start with the rows that can change your protocol: context and file capacity, plan usage, training controls, administrative access, and price. The matrix was checked on October 2, 2026; each figure identifies its vendor source and read date, while an unconfirmed cell is marked “Not verified” and left blank. Do not fill a missing business price, app-context figure, training setting, or advertising detail from memory. Check the linked vendor page before purchasing or deploying.