Lettertrace Review (2026): Pricing, Features & Honest Verdict

Lettertrace is an open-source AI visibility tracker that measures how often Claude, ChatGPT and Gemini mention your brand. It is free end to end because you bring your own API keys and pay the model providers directly. Best for: technical marketing teams who want AEO data without a $189/mo subscription. Price: Free (no paid tier). Rating: 7.5/10.

What is Lettertrace?

Lettertrace is a telemetry tool for a problem that barely existed two years ago: knowing whether AI assistants recommend your product when someone asks them for one. It runs a set of prompts against Claude, ChatGPT, Gemini and Google AI Overviews on a schedule, then reports how often you were mentioned, how prominently, in what tone, and how your share of voice compares to competitors you nominate.

We found it on Product Hunt’s 12 August board, where it ranked third with 56 upvotes. It comes from The Letter Company, with Mathew Pregasen credited as the lead maker alongside Deepak and Casey Millstein. The code lives at letterstory/lettertrace on GitHub under an MIT licence, created on 21 July 2026 and still being pushed to daily. At the time of writing the repo has 18 stars, which tells you exactly how early this is.

How does Lettertrace actually work?

The loop is four steps and the product does not hide any of them. You add the topics your buyers actually ask about, phrased the way they would phrase them: “best CRM for startups”, “email tools for agencies”. Lettertrace then expands each topic into dozens of natural prompt variations, so you are not hand-writing questions and accidentally testing only the ones you would win.

Each scheduled run fires those prompts at the models you have connected and parses the answers. Every run becomes a datapoint, so after a few weeks you have a trend line rather than a snapshot. That last part matters more than it sounds. A single AI answer is noisy, and any tool that shows you one result and calls it a ranking is selling you a coin flip.

What are Lettertrace’s key features?

Bring your own key, with no markup layer

This is the whole pitch. You plug in your own Anthropic, OpenAI and Google keys, and Lettertrace runs the queries through them. The keys are encrypted at rest. Because the platform never resells inference, there is no per-query credit to buy and no margin stacked on top of the model bill. Your only cost is what the providers charge you, which for a weekly run across a couple of dozen prompts is small money.

Multi-model coverage including AI Overviews

Claude, ChatGPT and Gemini are monitored side by side, plus Google AI Overviews. Covering AI Overviews is the useful inclusion, because it is the surface where AI answers and traditional search traffic actually collide. Plenty of AEO tools track chat assistants only and leave the biggest distribution channel out of the report.

Prompt variation generation

You give it a topic, it writes the questions. This removes the single biggest source of self-deception in AI visibility tracking, which is a marketer choosing the twelve prompts their product happens to answer well. Automated variation does not eliminate the bias entirely, since you still pick the topic, but it widens the aperture considerably.

Share of voice and sentiment, not just presence

The report separates four metrics: visibility, share of voice, prominence and sentiment. Being mentioned is not the same as being recommended, and being recommended third in a list of five is not the same as leading it. Lettertrace’s own demo makes this concrete, showing a sample brand at 41% share of voice against Notion at 28% and Linear at 19%, with a sentiment score of +0.34.

Scheduled runs that build a trend line

Runs can be set to daily or weekly on autopilot. This is the feature that turns a curiosity into a measurement, because a single query against a model is a sample of one and tells you almost nothing about how you are positioned. Weekly runs across a fixed prompt set give you a series you can plot, notice movement in, and correlate against the content you shipped that month.

Self-hosting

MIT licence, fork it, run it on your own Supabase. For a category where the data is a log of your own brand’s weak spots, being able to keep the whole thing inside your infrastructure is a real argument, not a checkbox.

How much does Lettertrace cost?

Nothing. There is no paid tier, no seat price, no credit pack, and no upgrade prompt waiting three screens in. You pay Anthropic, OpenAI and Google directly for the tokens your runs consume, and you pay Lettertrace zero.

That is a violent price for this category. Otterly.AI, one of the more transparent competitors, charges $29/month for a Lite plan capped at 15 search prompts, $189/month for 100 prompts on Standard, and $489/month for 400 prompts on Premium. Profound starts at $99/month and its popular tier is $399/month. Peec AI publishes a pricing page with no figures visible without interaction. Against a market where 100 tracked prompts costs $189/month, a free tool that generates dozens of variations per topic and charges you only for the underlying tokens is not competing on price, it is refusing to have the conversation.

The honest caveat: free software with no revenue model has a survival question attached. MIT licensing and self-hosting are the answer to that, but only if you actually fork it.

Who is Lettertrace best for?

Use Lettertrace if you have someone technical on the marketing side. Installing a CLI, getting three API keys and connecting a Supabase instance is not hard, but it is not nothing either, and a non-technical content lead will stall at step one. If your team already runs its own scripts and lives in a terminal, the setup cost disappears and you get the same category of data the $189/month tools sell.

Use it if your prompt volume is high. The paid tools price by tracked prompts, so the more topics you want to watch, the worse their arithmetic looks and the better BYOK looks.

Skip Lettertrace if you need this to be someone else’s problem. There is no account manager, no SLA, and no support queue behind an 18-star repo three weeks old. Skip it if your organisation cannot get API keys issued without a procurement cycle, because then the “free” tool costs you six weeks. And skip it if you need historical data today: every tracker starts its trend line the day you install it, and the paid incumbents cannot backfill that for you either, but they will at least hold your hand while you wait.

Best Lettertrace alternatives

Otterly.AI is the closest paid comparison and the most transparent about what you get. $29/month buys 15 search prompts, $189/month buys 100, and $489/month buys 400. If you want a hosted dashboard, someone to email, and a hard cap you can budget against, this is the trade Lettertrace asks you to give up.

Profound sits at the enterprise end, starting at $99/month with its highlighted plan at $399/month. It is the pick when AI visibility reporting has to land in front of a CMO in a format nobody argues with, and when you want conversation analytics beyond raw mention counts.

Peec AI targets marketing teams and SEO agencies specifically, with an agency-shaped feature set. Its pricing page renders no figures in the page source, so budget on the basis of a demo call rather than a published number.

What Lettertrace gets wrong

Two things. First, the category is crowded and getting more so, and a free entrant does not automatically win it: the paid tools have months of accumulated trend data and integrations into reporting stacks that Lettertrace has not built. Second, BYOK shifts a real operational burden onto you. Rate limits, key rotation, spend caps and provider outages all become your problem, and the tool that promised you no middleman is also the tool with no middleman to call when Gemini starts refusing your requests.

There is also a subtler issue with the whole category that Lettertrace does not solve: model answers vary run to run, and a share-of-voice figure computed from 24 answers has wide error bars nobody publishes. Lettertrace’s scheduled runs at least accumulate samples over time, which is the right shape of fix, but read any single week’s number with scepticism.

Final verdict: is Lettertrace worth it?

Yes, if you can run a CLI. AI visibility tracking is a category where the incumbents charge $189 to $489 a month to run prompts against APIs you could call yourself, and Lettertrace calls that out by doing exactly that and charging nothing. The metrics it produces, visibility, share of voice, prominence and sentiment, are the same four numbers the paid dashboards report. The methodology, automated prompt variation with scheduled repeat runs, is the correct one.

Rated 7.5 rather than higher because of maturity, not design. This is a three-week-old repo with 18 stars, no support channel, and a setup path that assumes competence you may not have on the marketing team. If that describes you, buy Otterly and get on with your day. If it does not, install Lettertrace, point it at your five most commercially important topics, and let it run for a month before you form any opinion about your AI visibility at all.