Quick Answer
Price anchoring is the pull a first reference price has on every price a buyer judges after it, and in AI SaaS pricing that reference is usually ChatGPT Plus at $20 a month. You beat it by changing what the buyer prices: charge for a finished job, cost out the manual work it replaces, and say both on your pricing page before the first call, rather than defending your inference bill.
The objection arrives early and plainly. "ChatGPT is $20 a month, so why does this cost so much?" It usually comes from a friendly prospect, sometimes a former colleague, on the first call.
That sentence is rarely about budget. The buyer is comparing a general chatbot they already understand with a product built to finish one job inside their team's workflow, and they have no other number to reach for.
Founders who answer it with a discount train every later buyer to expect one. This covers where the $20 reference comes from, why cost explanations fail, the five steps that move the price onto the job, and when the comparison is telling you something true.
Why Price Anchoring Pulls AI SaaS Toward $20
Price anchoring pulls AI SaaS toward $20 because ChatGPT Plus is the AI price a buyer is most likely to know already. A buyer with no reference for your category borrows the nearest one, and the nearest one is a consumer subscription.
OpenAI's own ChatGPT plan pricing shows how low and how memorable those numbers are. As of September 2026, the individual plans read like this.
| ChatGPT Plan | Monthly Price | Who It Is For |
|---|---|---|
| Free | $0 | Anyone trying it |
| Go | $8 | Lighter individual use |
| Plus | $20 | Regular individual use |
| Pro | From $100 | Heavy individual use |
None of those rows prices a finished piece of work. Each one prices access to a general assistant, where the user still writes the prompt, checks the output and carries it into the rest of the workflow.
That's the mismatch the anchor hides. Your buyer is holding a price for a tool they operate themselves and applying it to a product that operates for them.
Price Anchoring Hides That Your Product Sells a Different Unit
Your product and ChatGPT Plus are priced on different units, and price anchoring works only while the buyer can't see that. One sells a month of access to a model, and the other should sell a job completed.
| Factor | ChatGPT Plus | Your AI Product |
|---|---|---|
| What is bought | Access to a general model | One job finished in a workflow |
| Who does the work | The user, prompt by prompt | The product |
| Unit of value | A month of access | A ticket, letter or task done |
| What it replaces | Another chat tool | Hours of manual work |
Many founders see the gap and explain it through cost instead. They list tokens, retrieval, evaluation and support, which is honest, and it usually backfires.
The cost is real. Bessemer's AI pricing and monetization playbook reports that AI companies see 50 to 60 percent gross margin against 80 to 90 percent for classic SaaS, because every query carries compute cost.
But cost-plus pricing asks the buyer to price your infrastructure, and a buyer can't value your infrastructure. The same playbook names cost-plus pricing as a common mistake that leaves money on the table, and warns that soft return on investment kills willingness to pay.
How to Price AI SaaS on the Job Instead of the ChatGPT Anchor
Price your AI SaaS on the job it finishes, not the intelligence it accesses. Price anchoring to ChatGPT then loses its grip, because the buyer is no longer pricing the same thing. Five steps get you there, and each one feeds the next.
Jacob Jackson, co-founder of Supermaven, put the principle plainly in the Bessemer playbook: "The right way to price is relative to the value being delivered." The steps below turn that sentence into a number.
1. Name the One Job Your Product Finishes
Write the job as a noun your buyer already counts. A resolved support conversation, a drafted demand letter, a reconciled invoice or a qualified lead are jobs; "AI-powered insights" is not.
The test is whether the buyer's team tracks it today, in a spreadsheet, a queue or a weekly report. If they do, you have a unit they can price without learning anything new about AI.
You'll know it worked when a prospect repeats the unit back to you in their own words on the next call. The common mistake is naming three jobs at once, which sends the buyer back to the only single number they have, which is $20.
2. Choose a Charge Metric That Counts Finished Work
Pick the metric that bills for the job you named, not for the tokens underneath it. Bessemer sorts AI charge metrics into three families, and each one moves the buyer a different distance from the chatbot comparison.
| Charge Metric | What the Buyer Pays For | Distance From the $20 Anchor |
|---|---|---|
| Usage-based pricing | Tokens, calls or credits | Close, still priced as access |
| Workflow-based pricing | A task completed | Further, priced as work |
| Outcome-based pricing | A result achieved | Furthest, priced as a result |
Intercom Fin is a well-documented public example of outcome-based pricing. Intercom's pricing page charges $0.99 per Fin outcome, counted when a customer confirms the issue is resolved, asks for no more help after Fin responds, or Fin completes a workflow.
Bessemer's playbook describes EvenUp charging per completed personal injury demand letter rather than by the hour. In both cases the buyer compares the price with the cost of the work, and ChatGPT Plus never enters the sum.
Outcome pricing has a cost. You carry the variance when one outcome takes far more compute than another, so usage-based pricing or a base fee with outcome tiers is the safer start if you can't yet predict that spread.
3. Price the Manual Workflow With the Buyer's Own Numbers
Before you quote, find out what the job costs the buyer today. Ask three questions on the discovery call: how many of these jobs happen a month, how long each one takes, and who does it.
Multiply those answers together using the loaded hourly cost of that role, which the buyer can estimate faster than you can. The result is their reference price, and it comes from their own team rather than from your slide.
It worked if the buyer starts comparing your price with that figure instead of with a subscription. The mistake is filling in the numbers yourself, because an estimate the buyer didn't give is an estimate the buyer won't believe.
4. Find the Price Through Friction, Not Spreadsheets
Set a starting price, then raise it deal by deal until buyers pause. Bessemer's version is simple: if customers say "sold" immediately you're probably too cheap, so raise incrementally until you hear "we have to think about that," and stop just before price becomes a real blocker.
The discipline is in reading which objection you're hearing. A buyer who accepts the job framing but hesitates on the number has a packaging or proof problem, while a buyer who keeps returning to ChatGPT has a positioning problem.
Keep a note of the exact words each prospect uses when they push back. The mistake is treating every pause as a price problem and discounting, which fixes neither of the two real causes.
5. Put the Job and Its Cost on Your Pricing Page First
Your pricing page should state the unit of value and the manual cost it replaces above the number, so the buyer meets your reference before they reach for ChatGPT's. A pricing page that opens with tiers and feature checkmarks leaves price anchoring to whatever the buyer brought with them.
Your website is usually the first place a buyer checks you out, which is why Better Marketing builds sites that explain what you actually built. On your pricing page, that means one line naming the job, one line on what that job costs done by hand, then the price.
The wording is often already sitting in your support tickets and lost-deal notes, and this guide to rewriting a pricing page in your buyers' words walks through mining them. Most content marketing for tech companies stops at features, so a short page that prices the manual workflow is rare enough to be remembered.
You'll see it working when prospects arrive at the first call already quoting the unit. The mistake is burying the comparison in a blog post nobody reads before the demo.
What to Say When Price Anchoring Surfaces on a Sales Call
When price anchoring to ChatGPT comes up, move the buyer from comparing tools to comparing costs of the job. Don't argue that your model is smarter, because that keeps the conversation inside the chatbot's category.
Three lines do most of the work, and each one fits a different version of the objection:
- "ChatGPT is a general tool you operate yourself. We're not replacing a chatbot, we're replacing the manual work your team does around this task."
- "If your team only wants access to a model, ChatGPT Plus is probably enough. If you want this job finished inside your workflow, that's what we charge for."
- "The question isn't what a chatbot costs. It's what this work costs you today, in hours, delay and errors."
Say the second one and mean it. Conceding that ChatGPT fits some buyers makes the rest of your pricing believable, and it keeps you from discounting to win a buyer who never needed the product.
What AI Assistants Tell Buyers Your Product Costs
Price anchoring now starts before the call, because buyers ask ChatGPT and Perplexity what your product costs and whether it's worth it compared with ChatGPT itself. If your pricing page doesn't state a price and a unit, the assistant can fill the gap from review sites, forum threads and competitors' comparison pages.
That gap is where competitors end up owning the answer about your pricing. A number you never published still appears in the answer, and the unit attached to it is rarely the job you want priced.
If you pay for an AI SEO service, ask it to check what the main assistants say your product costs, not only where your pricing page ranks. An AI SEO service that reports rankings but never reads those answers is missing the place your first price is now set.
The same applies to anyone booking calls for you. A B2B demand generation agency that fills your calendar without your unit of value sends you prospects who arrive holding $20. Content marketing for tech companies that never states the unit leaves that agency nothing to point to.
When Price Anchoring to ChatGPT Is Telling You the Truth
Sometimes price anchoring to ChatGPT is correct, because the buyer really could do the job with a subscription and a few prompts. When that happens, the problem is the use case or the customer, and no pricing move fixes it.
You'll recognize it when the buyer describes the job as occasional, low stakes and done by one person. A task someone does twice a month in a chat window has no manual workflow to cost out, so there's nothing for the job framing to attach to.
You have two honest options. Narrow the product to a job that is repetitive, regulated or spread across a team, where a general chatbot clearly falls short, or move to a buyer whose version of the job is expensive enough to price. Don't spend months defending a premium price on a use case ChatGPT Plus already covers.
Write the Unit of Value Down Before the Next Call
Price anchoring to ChatGPT fades once the buyer prices a finished job instead of a chatbot. Write the one job your product finishes, the three numbers that cost it out, and the pricing page line that states both, then test the price through friction rather than discounting.
If you'd rather build that trust before a buyer reaches the pricing page, Better Marketing works as a B2B demand generation agency that earns it through founder-led content and outreach, without paid ads.