https://forthsay.com/whats-new/ 2026-11-07T17:00:00Z Forthsay https://forthsay.com/whats-new/#own-ads-count2026-11-07T17:00:00Z

Our anonymous daily counts now tell one more kind of source apart: a visit from an ad of our own, known only by a tag in that ad’s link. A link someone shares on a social site still counts as another site. The privacy page says so. We still run no ad pixel and no tracking cookie, and nothing more about you is counted.

https://forthsay.com/whats-new/#input-feedback2026-11-07T17:00:00Z

The sign-in page now answers whatever you do: the switch slides between signing in and a new account, and the heading and button change with it. Your email gets a tick once it reads as an address. The password has a show button, a Caps Lock warning and, for a new account, a bar that fills as it gets long enough. The button spins while it works and ticks when you are in, and a wrong password is marked so you can retype it at once. On the dashboard the picked tab slides to your new pick, bars grow and move to their new values, numbers count to them, the trend line draws itself, the answer squares fill in when you open Answers, and the light and dark switch fades across. None of it plays if your device asks for less motion.

https://forthsay.com/whats-new/#case-pages2026-11-07T17:00:00Z

Makerfern and Cuescribe, our founder’s own sites, each have a page now. Each shows the site’s own records (Makerfern’s ChatGPT visitors day by day, and where each site’s first visits came from as 100 squares), all 32 answers from our 7 Nov check as a grid of squares, each engine’s share with its 95% range and the sites it cited instead, the two lines where ChatGPT recommended Makerfern word for word, and what a plan would track for that site. On the home page the three sites sit side by side with their own logos: Makerfern’s chart, Cuescribe’s sources and Pencilhop’s crawler numbers, the new site we are watching as a live test.

https://forthsay.com/whats-new/#free-dashboard2026-11-07T17:00:00Z

A free account now opens the dashboard too, with no plan needed. It shows your one free brand check the way a plan shows its weeks: your recommended, named and cited shares with their 95% ranges, the dial, each engine, every answer word for word and the four questions. While the check runs, the dashboard counts its answers in. The parts that come with a plan (the week-by-week chart, analytics, rivals, changes, your own questions, the facts file, sharing, exports and the API) sit behind a slowly moving field of squares, with our made-up sample brand blurred underneath and a line on what a plan adds there. The account button in the top bar is now called Dashboard, the free check links to it once you are signed in, and forthsay.com/dashboard/ opens it too.

https://forthsay.com/whats-new/#llms-method2026-11-07T17:00:00Z

Our llms.txt file, the page AI assistants read about a site, now holds our method in full: where every answer comes from (each vendor’s official API with web search on, never the apps), the model behind each engine, how often we ask, the rules for named, recommended and cited, the 95% ranges, how a change gets its verdict, and why answers that did not come back are left out. Our facts file names the four models, and robots.txt points to llms.txt.

https://forthsay.com/whats-new/#report-dashboard2026-11-07T17:00:00Z

Your account, the sample report and shared links now open on one dashboard with pages for Overview, Analytics, Answers, Rivals, Changes and Questions; your account adds Settings. The overview has your recommended, named and cited shares with this week’s 95% range, a dial, a week-by-week chart with your change markers, each engine and your rivals. Analytics draws any share over 4 weeks, 8 weeks or all of them, beside the period before, with the raw numbers week by week. Answers lays out every answer of the latest week as squares; tap one to read it word for word, with the line that decided it. Changes is a board: what is worth a look, changes still being measured, and changes with a verdict. Every number is still k of n answers, and answers that did not come back are left out of n. The sample report now runs ten weeks of data, and a search-setup step no longer breaks a bold word onto a line of its own.

https://forthsay.com/whats-new/#grading-pick-lines2026-11-07T17:00:00Z

We checked our method against our founder’s own sites, whose visitors come mostly from ChatGPT. ChatGPT often picks a site with lines like “try Makerfern’s free pattern maker” or “Easiest if you don’t want to install anything: Makerfern”, and the grader read those as a passing name. They now count as recommendations, and so does “Best free option overall: …”. “Don’t use …” still does not. When an engine blocks a question for copyright, the answer is now marked refused rather than an error.

https://forthsay.com/whats-new/#gemini-search-model2026-11-07T17:00:00Z

Gemini answers came back as errors in some early free checks. Gemini now runs on Gemini 3.5 Flash-Lite, which Google lists as able to search the web while it answers, and a live test on 7 Nov got answers with sources. Gemini’s sources now count too: Google sends its own redirect links, so a Gemini answer that cited your site was not counted as citing you; we now read the site’s name that comes with each link. When any engine sends no answer, the free check and your account now say why in plain words (a rate limit, a refused key, a model the engine does not offer, a server fault or a timeout). Those answers are still left out of every share.

https://forthsay.com/whats-new/#perplexity-rate-limit2026-11-07T17:00:00Z

When Perplexity asks us to slow down briefly, we wait before retrying, within the original answer time limit. Longer rate limits and other failures remain not measured.

https://forthsay.com/whats-new/#turnstile-verification2026-11-07T17:00:00Z

The free brand check verifies that its Turnstile token belongs to this site and this action before it starts. Expired, reused or mismatched tokens are refused. The report’s unbranded summary now counts only unbranded failures, so its measured and unmeasured counts add up.

https://forthsay.com/whats-new/#logo2026-11-07T17:00:00Z

Forthsay’s logo is now an opening quote mark, because every report shows what the AI said, word for word. It is in the top bar, the footer, the browser tab and the icon your phone saves.

https://forthsay.com/whats-new/#own-sites2026-11-07T17:00:00Z

The method page now shows why each rule exists, with the records behind it: Makerfern’s ChatGPT visitors day by day (from about 42 a day in late September to about 132 by late October), what Cuescribe taught us about Bing, Google and old prices, and Pencilhop, a new site we are watching as a live test. You can paste any AI answer into the method page and see how the rules grade it. The home page has the same three sites in short, the pricing page shows what every plan includes side by side with a questions section, the sign-in page has a new layout, and every page has a new footer.

https://forthsay.com/whats-new/#search-setup2026-11-07T17:00:00Z

The search setup check looks for public signs that a site is verified in Google Search Console and Bing Webmaster Tools, and finds and counts its sitemap. ChatGPT’s web answers draw on Bing’s index, so the free brand check report and each site in your account now run it by themselves and show the steps when something is missing. Some ways to verify leave no public trace, so “no sign” says to look in your account, never “not set up”.

https://forthsay.com/whats-new/#claude-prompts2026-11-07T17:00:00Z

The API and MCP page has two Copy prompt buttons. Paste one into Claude and it adds the MCP server itself (Claude Code runs the command, Claude Desktop edits its settings file, anywhere else it gives you the steps) and tries a tool, or it writes a script that saves your weekly numbers to a CSV. Inside your account the same prompts come with your API key and your sites filled in. You can read each prompt before you copy it.

https://forthsay.com/whats-new/#free-check-account2026-11-07T17:00:00Z

Each free brand check asks four AI engines 32 questions through their paid APIs, so it now runs once per account: make a free account with Google or an email and password, and the check page brings you straight back with what you typed. If someone already checked your domain in the last 30 days you get that report at once, and your own check stays unused. The other free tools still need no account.

https://forthsay.com/whats-new/#accounts2026-11-07T17:00:00Z

Plans now live on an account: sign in with Google or with an email and password, pick a plan on the pricing page, and the plan lands on your account as soon as Whop confirms the payment, with no key to paste. The app has an Account page with your API key for the REST API and the MCP server, a button to make a new key that stops the old one at once, sign-out on every device, and account deletion. Each site can have a read-only report link you turn on and off, and the answers download as CSV or JSON. The privacy page lists the three sign-in cookies; browsing and the free tools still set none.

https://forthsay.com/whats-new/#plans-more2026-11-07T17:00:00Z

Prices stay at $39, $99 and $249 a month. Every plan now has more sites and more rivals (Starter 2 sites and 8 rivals each, Growth 5 and 15, Scale 15 and 25), the REST API as well as the MCP server, read-only report links you can share, and CSV and JSON export. Each question is now asked twice a week on each engine instead of three times, which keeps the price where it is; the ranges on every number say how much that leaves you knowing.

https://forthsay.com/whats-new/#sample-desk2026-11-07T17:00:00Z

The sample report is now a dashboard: switch between recommended, named and cited, hover or tap any week on the chart, chart one engine alone, and see before-and-after ranges for a change. Every answer from the latest week sits in a grid you can filter and open, with the raw counts under each cell, a full table of k of n with 95% ranges, the range worked out by hand, and a JSON download. Across the site the Growth price is readable again, the top bar keeps one width on every page, the tool icons share one style, tool pages put the tool beside its limits, and reading pages use the full width on large screens. The sample report left the top bar; the home dashboard still links to it.

https://forthsay.com/whats-new/#pricing-note2026-11-06T17:00:00Z

While plans are not on sale yet, the note on the pricing page now names only the parts of the site that are switched on, so it no longer says the free brand check works before it does.

https://forthsay.com/whats-new/#free-check-live2026-11-06T17:00:00Z

The free brand check has a new layout. While it runs you see a grid of every engine and question filling in as each answer comes back, with the newest answer written out underneath. The finished report opens on your three shares with their 95% ranges, then a row per engine, then every answer as a grid you can tap to read that answer and its sources, with your name marked. It works the same on a phone. We also fixed two cut-off bits of text: the engine name in the home headline and long model names on the home dashboard.

https://forthsay.com/whats-new/#engine-roll2026-11-06T17:00:00Z

The headline on the home page now rolls through ChatGPT, Claude, Gemini and Perplexity, each with its mark, so you can see which engines we ask.

https://forthsay.com/whats-new/#live-dashboard2026-11-06T17:00:00Z

The home page now opens on a sample dashboard: an eight-week trend with its 95% band, rivals on the same questions, a row per engine with its model, and a stored answer being written out with its verdicts. Each engine we ask is shown by name and mark. Type your domain on the home page and the free brand check opens with it filled in. Across the site, fields now say when what you typed is usable, buttons show when they are working, and copies and toggles confirm themselves.

https://forthsay.com/whats-new/#opening2026-11-06T17:00:00Z

The tracker asks the models behind ChatGPT, Claude, Gemini and Perplexity your questions every week, three times each, and shows how often you were named, recommended and cited with a 95% range on every number and the raw answer behind it. Twelve free tools come with it, including a one-time brand check across four engines, a bot view, a robots reader by bot type, a facts file builder and a crawler log verifier. The method page prints every rule, and the risks page explains the practices we will not use.