Did AI Subscription Prices Really Rise? 2023–2026

The single most durable number in consumer AI is $20 a month. ChatGPT Plus set it on 1 February 2023 and has never raised it. Claude Pro arrived on 7 September 2023 at exactly the same price and has never raised it either. Perplexity Pro has held around $20 since 2023, and when Google finally entered the paid market in February 2024 it priced Google One AI Premium at $19.99 — a cent under the incumbent rather than a challenge to it.

That is unusual. Consumer software subscriptions normally drift upward, and the underlying product here changed beyond recognition over the same period: context windows grew, reasoning modes appeared, and the models behind the same $20 plan were replaced several times over.

Price Anchoring · $20 Across Four Companies

The consistency is not coincidence so much as anchoring. Once ChatGPT Plus established $20 as the price of a flagship chatbot, matching it became the safe move and undercutting it signalled a weaker product. Anthropic matched to the dollar. Google came in a cent below. Perplexity, selling a different product — a cited answer engine rather than a chatbot — landed on the same number anyway. The figure stopped being a pricing decision and became a category convention.

Only xAI broke ranks upward. Standalone SuperGrok launched around February 2025 at $30, the one entry tier in this comparison priced above the norm. Microsoft, meanwhile, matched at $20 in January 2024 and then walked its consumer plan back entirely — the only provider whose entry price moved down rather than staying put.

In 2023 the ceiling question had no answer, because there was no ceiling. The standard plan was the most expensive consumer option every provider sold, which is why the floor-to-ceiling spread for that year is exactly 1x across the board. Paying more was not an available choice.

ChatGPT Pro changed that on 5 December 2024 at $200 a month — ten times the price of the plan directly beneath it, and the first consumer AI tier at that level. The number was widely treated as an outlier at launch.

The Cascade · Five Providers in Eight Months

It did not stay an outlier. Anthropic launched Claude Max on 9 April 2025 at $100 and $200, framed explicitly as an answer to ChatGPT Pro. Google put AI Ultra at $249.99 on 20 May 2025, the highest opening price of the group. Perplexity Max reached $200 by mid-2025. xAI closed the run on 9 July 2025 with SuperGrok Heavy at $300, offered at $99 for a subscriber's first three months. A bracket that was empty in November 2024 held five products by the following summer.

What these tiers sell is capacity rather than capability. The $200 plan is generally the same model family as the $20 plan with roughly twenty times the usage allowance, which is why the pricing scales so cleanly and why the tiers are aimed at a narrow group of heavy users — people running long agentic sessions or working inside the tools all day — rather than at the general subscriber.

The sharpest finding in this dataset is not the size of the spread but how similar the ladders have become. By August 2026 five providers sell the same three-rung structure: roughly $20 for the standard plan, about $100 for five times the usage, about $200 for twenty times it. The prices line up almost to the dollar, and so do the multipliers.

Opposite Directions · Google Cut, OpenAI Added

They did not arrive there together. Google launched its ceiling highest, at $249.99, and then cut and restructured it into a $99.99 five-times tier and a $199.99 twenty-times tier — pulling its top price down into the pack rather than pushing above it. OpenAI moved the other way, adding a $100 rung beneath its existing $200 Pro on 9 April 2026 to fill the gap Anthropic had occupied since 2025. One company came down, another filled in, and the two menus ended up mirroring each other.

That convergence is worth reading carefully. Identical price points across independent competitors usually signal that everyone is solving the same constraint — here, the cost of serving heavy inference to a small number of very active users — rather than that anyone found a pricing edge. The $100 and $200 rungs are priced off usage economics, not off what the market will bear.

It also means the ceiling is no longer climbing. The most recent moves in this dataset are a cut and a fill, not a raise. Whether $200 turns out to be a durable upper bound or a way station is the open question the next twelve months will answer.

xAI is the only provider in this comparison that sits above the field at both ends. Standalone SuperGrok costs $30 against a $20 norm, and SuperGrok Heavy costs $300 against a $200 one — 50% higher at the top and 50% higher at the bottom. Its internal spread is still exactly 10x, the same ratio as everyone else; the whole ladder is simply shifted upward.

Grok also reached the market by a different route. It shipped first as a bundled feature inside X Premium+ in November 2023 rather than as a standalone subscription, and the standalone tier did not appear until roughly February 2025 — well after the $20 convention had hardened elsewhere.

The $300 Tier · Highest Consumer Ceiling in the Market

SuperGrok Heavy, launched 9 July 2025, remains the most expensive individual AI subscription of the six providers here. The $99 introductory rate for a subscriber's first three months is itself a signal: a third of list price is a steep enough discount to suggest the headline number is doing positioning work as much as revenue work. Measured against the field's $20 entry norm rather than xAI's own $30 floor, that ceiling sits 15x above the price most people pay.

Microsoft is the counterexample that makes the rest of the pattern legible. Copilot Pro launched in January 2024 at $20, squarely on the convention. Then Microsoft went the other way: the standalone plan was withdrawn from new buyers on 1 October 2025 and folded into Microsoft 365 Premium at $19.99, with existing Copilot Pro subscriptions set to stop renewing on 1 August 2026.

More striking is what never appeared. Microsoft has no $200-class consumer chat tier and never built one. Its high end is Copilot Studio at $200 a month, which is an agent-building add-on rather than a power-user chat plan, and enterprise Microsoft 365 Copilot at $30 per seat, which is a different buyer entirely.

Why the Absence Matters · A Choice, Not an Inevitability

Every other provider treated the $200 tier as a necessary response once OpenAI created it. Microsoft, with more consumer distribution than any of them, decided the opposite — that AI belongs bundled into a $19.99 productivity subscription rather than sold as a premium standalone. That single refusal is the strongest evidence in this dataset that the $20-to-$200 ladder was a strategic choice the industry made, not an economic law it discovered.

While attention went to the top of the ladder, the more consequential change for most people happened at the bottom. A set of tiers now sits beneath the old $20 floor: ChatGPT Go at $8, launched globally in January 2026 after an India pilot; Google AI Plus at $4.99; SuperGrok Lite at $10; and Perplexity Education Pro at $10 for verified students.

This is the detail that most directly contradicts the assumption that AI subscriptions keep getting more expensive. The entry price did not rise. The market added rungs underneath it, which means the cheapest way to pay for a capable AI assistant in 2026 costs a quarter of what the cheapest paid option cost in 2023.

The trade is usually usage limits, model access, or in ChatGPT Go's case advertising. But the direction of travel at the volume end of the market is unambiguously downward, and it is happening at the same time as the $200 tiers that generate the headlines.

This comparison measures list prices, and list prices hide a great deal. A $200 tier is not ten times more capable than a $20 one; it is generally the same models with roughly twenty times the usage allowance. Anyone reading the spread as a capability gap is reading it wrong.

Bundles and Billing · Not Like-for-Like

Several of these plans are not pure AI subscriptions at all. Google's tiers include cloud storage and other consumer services; Microsoft's $19.99 buys the Office applications alongside the AI features. Annual billing also cuts the real price — Claude Pro drops to $17 a month when paid yearly, and Perplexity Pro to roughly $16.67 — so the monthly headline numbers overstate what committed subscribers actually pay.

Two further limits are worth stating plainly. These are US list prices, and regional pricing diverges sharply — the same tiers cost very different amounts in local currency elsewhere. And the comparison says nothing about value per dollar: the $20 that bought GPT-4-era access in 2023 buys a far more capable system in 2026, which is arguably a real-terms price cut that a nominal-price chart cannot show.

The dataset is also perishable. Three of the most significant figures here — ChatGPT Go, the $100 Pro rung, and Google's Ultra restructure — all landed within the last eight months, and any of them could move again before this is a year old.

So did AI subscription prices rise between 2023 and 2026? For the roughly $20 plan that most subscribers buy, no — it is the same price it was on 1 February 2023, and cheaper options now exist beneath it that did not exist then.

For the top of the market, emphatically yes. A bracket that was empty in 2023 now runs to $200 across five providers and $300 at xAI, which is a real and large increase in what it is possible to spend on a consumer AI subscription.

Both things are true at once, and conflating them is what makes the popular complaint feel right while being wrong about the number that matters to most people. The ceiling rose. The floor did not. The distance between them is the actual story, and at 10x it is wider than in almost any other consumer software category.

The direction of travel from here looks flatter than the last two years suggest. Google cut its ceiling rather than raising it, OpenAI's most recent move was to add a cheaper rung, and the newest tiers in the market are the $4.99 and $8 ones. If the pattern holds, the interesting pricing action has already moved from the top of the ladder to the bottom.