New analysis exhibits that the best way AI providers invoice by tokens hides the true price from customers. Suppliers can quietly inflate costs by fudging token counts or slipping in hidden steps. Some techniques run additional processes that don’t have an effect on the output however nonetheless present up on the invoice. Auditing instruments have been proposed, however with out actual oversight, customers are left paying for greater than they understand.
In almost all circumstances, what we as customers pay for AI-powered chat interfaces, equivalent to ChatGPT-4o, is presently measured in tokens: invisible models of textual content that go unnoticed throughout use, but are counted with actual precision for billing functions; and although every trade is priced by the variety of tokens processed, the person has no direct method to affirm the rely.
Regardless of our (at finest) imperfect understanding of what we get for our bought ‘token’ unit, token-based billing has change into the usual strategy throughout suppliers, resting on what could show to be a precarious assumption of belief.
Token Phrases
A token just isn’t fairly the identical as a phrase, although it typically performs an identical function, and most suppliers use the time period ‘token’ to explain small models of textual content equivalent to phrases, punctuation marks, or word-fragments. The phrase ‘unbelievable’, for instance, is perhaps counted as a single token by one system, whereas one other would possibly cut up it into un, believ and ready, with every bit rising the price.
This technique applies to each the textual content a person inputs and the mannequin’s reply, with the value primarily based on the full variety of these models.
The problem lies in the truth that customers don’t get to see this course of. Most interfaces don’t present token counts whereas a dialog is occurring, and the best way tokens are calculated is difficult to breed. Even when a rely is proven after a reply, it’s too late to inform whether or not it was truthful, making a mismatch between what the person sees and what they’re paying for.
Latest analysis factors to deeper issues: one research exhibits how suppliers can overcharge with out ever breaking the foundations, just by inflating token counts in ways in which the person can not see; one other reveals the mismatch between what interfaces show and what’s really billed, leaving customers with the phantasm of effectivity the place there could also be none; and a third exposes how fashions routinely generate inner reasoning steps which can be by no means proven to the person, but nonetheless seem on the bill.
The findings depict a system that appears exact, with actual numbers implying readability, but whose underlying logic stays hidden. Whether or not that is by design, or a structural flaw, the end result is similar: customers pay for greater than they will see, and sometimes greater than they anticipate.
Cheaper by the Dozen?
Within the first of those papers – titled Is Your LLM Overcharging You? Tokenization, Transparency, and Incentives, from 4 researchers on the Max Planck Institute for Software program Programs – the authors argue that the dangers of token-based billing prolong past opacity, pointing to a built-in incentive for suppliers to inflate token counts:
‘The core of the issue lies in the truth that the tokenization of a string just isn’t distinctive. For instance, take into account that the person submits the immediate “The place does the following NeurIPS happen?” to the supplier, the supplier feeds it into an LLM, and the mannequin generates the output “|San| Diego|” consisting of two tokens.
‘For the reason that person is oblivious to the generative course of, a self-serving supplier has the capability to misreport the tokenization of the output to the person with out even altering the underlying string. For example, the supplier may merely share the tokenization “|S|a|n| |D|i|e|g|o|” and overcharge the person for 9 tokens as a substitute of two!’
The paper presents a heuristic able to performing this sort of disingenuous calculation with out altering seen output, and with out violating plausibility underneath typical decoding settings. Examined on fashions from the LLaMA, Mistral and Gemma sequence, utilizing actual prompts, the strategy achieves measurable overcharges with out showing anomalous:
Token inflation utilizing ‘believable misreporting’. Every panel exhibits the share of overcharged tokens ensuing from a supplier making use of Algorithm 1 to outputs from 400 LMSYS prompts, underneath various sampling parameters (m and p). All outputs had been generated at temperature 1.3, with 5 repetitions per setting to calculate 90% confidence intervals. Supply: https://arxiv.org/pdf/2505.21627
To handle the issue, the researchers name for billing primarily based on character rely reasonably than tokens, arguing that that is the one strategy that offers suppliers a purpose to report utilization truthfully, and contending that if the purpose is truthful pricing, then tying price to seen characters, not hidden processes, is the one possibility that stands as much as scrutiny. Character-based pricing, they argue, would take away the motive to misreport whereas additionally rewarding shorter, extra environment friendly outputs.
Right here there are a selection of additional issues, nevertheless (typically conceded by the authors). Firstly, the character-based scheme proposed introduces extra enterprise logic that will favor the seller over the buyer:
‘[A] supplier that by no means misreports has a transparent incentive to generate the shortest potential output token sequence, and enhance present tokenization algorithms equivalent to BPE, in order that they compress the output token sequence as a lot as potential’
The optimistic motif right here is that the seller is thus inspired to supply concise and extra significant and precious output. In observe, there are clearly much less virtuous methods for a supplier to scale back text-count.
Secondly, it’s cheap to imagine, the authors state, that firms would doubtless require laws to be able to transit from the arcane token system to a clearer, text-based billing methodology. Down the road, an rebel startup could determine to distinguish their product by launching it with this sort of pricing mannequin; however anybody with a really aggressive product (and working at a decrease scale than EEE class) is disincentivized to do that.
Lastly, larcenous algorithms such because the authors suggest would include their very own computational price; if the expense of calculating an ‘upcharge’ exceeded the potential revenue profit, the scheme would clearly don’t have any advantage. Nevertheless the researchers emphasize that their proposed algorithm is efficient and economical.
The authors present the code for his or her theories at GitHub.
The Swap
The second paper – titled Invisible Tokens, Seen Payments: The Pressing Have to Audit Hidden Operations in Opaque LLM Providers, from researchers at the College of Maryland and Berkeley – argues that misaligned incentives in industrial language mannequin APIs are usually not restricted to token splitting, however prolong to total courses of hidden operations.
These embrace inner mannequin calls, speculative reasoning, instrument utilization, and multi-agent interactions – all of which can be billed to the person with out visibility or recourse.
Pricing and transparency of reasoning LLM APIs throughout main suppliers. All listed providers cost customers for hidden inner reasoning tokens, and none make these tokens seen at runtime. Prices differ considerably, with OpenAI’s o1-pro mannequin charging ten occasions extra per million tokens than Claude Opus 4 or Gemini 2.5 Professional, regardless of equal opacity. Supply: https://www.arxiv.org/pdf/2505.18471
Not like standard billing, the place the amount and high quality of providers are verifiable, the authors contend that as we speak’s LLM platforms function underneath structural opacity: customers are charged primarily based on reported token and API utilization, however don’t have any means to verify that these metrics mirror actual or mandatory work.
The paper identifies two key types of manipulation: amount inflation, the place the variety of tokens or calls is elevated with out person profit; and high quality downgrade, the place lower-performing fashions or instruments are silently used instead of premium parts:
‘In reasoning LLM APIs, suppliers typically preserve a number of variants of the identical mannequin household, differing in capability, coaching knowledge, or optimization technique (e.g., ChatGPT o1, o3). Mannequin downgrade refers back to the silent substitution of lower-cost fashions, which can introduce misalignment between anticipated and precise service high quality.
‘For instance, a immediate could also be processed by a smaller-sized mannequin, whereas billing stays unchanged. This observe is troublesome for customers to detect, as the ultimate reply should still seem believable for a lot of duties.’
The paper paperwork situations the place greater than ninety % of billed tokens had been by no means proven to customers, with inner reasoning inflating token utilization by an element higher than twenty. Justified or not, the opacity of those steps denies customers any foundation for evaluating their relevance or legitimacy.
In agentic techniques, the opacity will increase, as inner exchanges between AI brokers can every incur costs with out meaningfully affecting the ultimate output:
‘Past inner reasoning, brokers talk by exchanging prompts, summaries, and planning directions. Every agent each interprets inputs from others and generates outputs to information the workflow. These inter-agent messages could eat substantial tokens, which are sometimes circuitously seen to finish customers.
‘All tokens consumed throughout agent coordination, together with generated prompts, responses, and tool-related directions, are sometimes not surfaced to the person. When the brokers themselves use reasoning fashions, billing turns into much more opaque’
To confront these points, the authors suggest a layered auditing framework involving cryptographic proofs of inner exercise, verifiable markers of mannequin or instrument id, and impartial oversight. The underlying concern, nevertheless, is structural: present LLM billing schemes depend upon a persistent asymmetry of knowledge, leaving customers uncovered to prices that they can not confirm or break down.
Counting the Invisible
The ultimate paper, from researchers on the College of Maryland, re-frames the billing drawback not as a query of misuse or misreporting, however of construction. The paper – titled CoIn: Counting the Invisible Reasoning Tokens in Business Opaque LLM APIs, and from ten researchers on the College of Maryland – observes that the majority industrial LLM providers now cover the intermediate reasoning that contributes to a mannequin’s last reply, but nonetheless cost for these tokens.
The paper asserts that this creates an unobservable billing floor the place total sequences could be fabricated, injected, or inflated with out detection*:
‘[This] invisibility permits suppliers to misreport token counts or inject low-cost, fabricated reasoning tokens to artificially inflate token counts. We check with this observe as token rely inflation.
‘For example, a single high-efficiency ARC-AGI run by OpenAI’s o3 mannequin consumed 111 million tokens, costing $66,772.3 Given this scale, even small manipulations can result in substantial monetary influence.
‘Such info asymmetry permits AI firms to considerably overcharge customers, thereby undermining their pursuits.’
To counter this asymmetry, the authors suggest CoIn, a third-party auditing system designed to confirm hidden tokens with out revealing their contents, and which makes use of hashed fingerprints and semantic checks to identify indicators of inflation.
Overview of the CoIn auditing system for opaque industrial LLMs. Panel A exhibits how reasoning token embeddings are hashed right into a Merkle tree for token rely verification with out revealing token contents. Panel B illustrates semantic validity checks, the place light-weight neural networks examine reasoning blocks to the ultimate reply. Collectively, these parts enable third-party auditors to detect hidden token inflation whereas preserving the confidentiality of proprietary mannequin conduct. Supply: https://arxiv.org/pdf/2505.13778
One part verifies token counts cryptographically utilizing a Merkle tree; the opposite assesses the relevance of the hidden content material by evaluating it to the reply embedding. This permits auditors to detect padding or irrelevance – indicators that tokens are being inserted merely to hike up the invoice.
When deployed in assessments, CoIn achieved a detection success fee of almost 95% for some types of inflation, with minimal publicity of the underlying knowledge. Although the system nonetheless is determined by voluntary cooperation from suppliers, and has restricted decision in edge circumstances, its broader level is unmistakable: the very structure of present LLM billing assumes an honesty that can not be verified.
Conclusion
In addition to the benefit of gaining pre-payment from customers, a scrip-based foreign money (such because the ‘buzz’ system at CivitAI) helps to summary customers away from the true worth of the foreign money they’re spending, or the commodity they’re shopping for. Likewise, giving a vendor leeway to outline their personal models of measurement additional leaves the buyer in the dead of night about what they’re really spending, by way of actual cash.
Just like the lack of clocks in Las Vegas, measures of this sort are sometimes aimed toward making the buyer reckless or detached to price.
The scarcely-understood token, which could be consumed and outlined in so some ways, is maybe not an acceptable unit of measurement for LLM consumption – not least as a result of it could actually price many occasions extra tokens to calculate a poorer LLM lead to a non-English language, in comparison with an English-based session.
Nevertheless, character-based output, as instructed by the Max Planck researchers, would doubtless favor extra concise languages and penalize naturally verbose languages. Since visible indications equivalent to a depreciating token counter would most likely make us just a little extra spendthrift in our LLM periods, it appears unlikely that such helpful GUI additions are coming anytime quickly – at the least with out legislative motion.
* Authors’ emphases. My conversion of the authors’ inline citations to hyperlinks.
First revealed Thursday, Might 29, 2025
