“What should I charge for a brand identity project?” is the most tempting freelance prompt and one of the worst. The model will return a confident range with tidy packages. That range is not your market, your costs, or your obligation. Sending it onward as if it were research is how underpricing and overpromising both get a clean UI.
This article teaches research hygiene. It will not tell you what to charge. If you want proposal structure after you have a number, use proposal draft you still price.
What counts as pricing research
Useful inputs you gather:
- Your cost floor (software, taxes set-aside, unpaid admin hours, insurance) - calculated by you or with an accountant, not guessed by a chat
- Time estimates from your own past similar work
- Public rate surveys or association data you can cite, with date and population noted
- Peer conversations (informational, not cartel behavior - do not coordinate prices with competitors)
- Client budget signals they volunteered
- Positioning: speed, specialty, risk, support level
Useless substitutes:
- A single model-generated “market rate”
- Anonymous forum screenshots with no date or scope match
- Competitor website packages scraped without knowing what is included
Never ask a model what you should charge and paste that number into a client proposal. If you cannot explain the fee from your costs, positioning, and evidence, it is not ready.
Workflow
Step 1: Write your decision inputs (human)
Complete:
- Role / offer definition (one sentence)
- What is included / excluded
- Your hours estimate and confidence
- Cost floor notes
- Strategic goal (book work / raise floor / flagship case)
Step 2: Source list with AI as organizer only
Help me organize a pricing research plan for this offer.
Do not suggest a price or price range.
Propose: source types to check, questions to ask peer advisors,
and a table template for recording date, source, scope match,
and notes. Offer definition: [your sentence]
Step 3: Fill the table yourself
For each source row: date accessed, what scope they describe, how well it matches your offer, and what you learned - without copying someone else’s fee as yours.
Step 4: Decide offline
Sleep on it. Pick a fee or a range you will defend on a call. Enter that number into your proposal spine by hand.
Step 5: Claims check if you publish packages
Public package pages are commercial communications. Keep them truthful and substantiated (FTC advertising and marketing; FTC crackdown on deceptive AI claims and schemes; EU UCPD).
Do not paste client budget spreadsheets, named prospect financials, or confidential win/loss notes into a consumer tool while “researching rates.”
Maintain a private rate journal: date, offer, fee chosen, win/loss, lesson. That journal beats any chatbot range after a few months of real data.
Illustrative scenario (labeled)
Illustrative scenario, not a measured case: A developer asks for “standard hourly rates for React freelancers in Europe, 2026.” The model returns a single band. The developer bids the midpoint on a high-risk integration with on-call expectations. The project consumes unpaid nights. The band never mentioned on-call. The research failure was scope match, not lack of a number.
Packages vs hourly vs value framing
How you present a fee is separate from how you researched it. AI can help you list package names and included bullets after you choose the numbers. It should not choose whether you are “premium” or invent competitor comparisons you did not verify. If you publish three packages, each package needs a scope you can deliver on a bad week, not only on an ideal week.
Peer conversations without collusion theater
Talking to trusted peers about how they think about cost floors and positioning can be educational. Coordinating prices with competitors to fix a market is a different legal territory entirely - do not use research prompts to draft price-fixing schemes. Keep peer notes as qualitative inputs in your journal, not as a shared rate sheet. Solo-operator resource hubs can point you to education without dictating a fee (Freelancers Union resources; SBA: manage your business).
Validation and fallback
Validation: your proposal fee appears in your rate journal with a one-line rationale that does not cite “ChatGPT said.”
Fallback: cost-plus from your floor and hours, then adjust for positioning - still your arithmetic.
Tax treatment of income is accountant territory (when solo AI is not enough). Contract payment terms are literacy plus professional review (contract invoice literacy).
