Making Discoverable discoverable.

A reading is never rewritten. When the method changes, the readings before and after it are different quantities and are shown in separate groups rather than on one line. The current method measures seven engines (Claude, ChatGPT, Perplexity, Gemini, Copilot, Google AI Overviews and Google AI Mode) from logged-out sessions at a UK location, counting a source only when the answer cited it. Method details are on the methodology page.

Every reading

Grouped by method. Within a group the readings are comparable; across groups they are not. Score is the product's visibility score on that method; named and cited are counts of answers.

What we changed, and when

Two kinds of entry. A site change is what a customer would do. A measurement change alters how the readings are taken and starts a new group above. Movement after a change is an association; other factors, including concurrent changes, may explain it.

1. 18 Jul 2026Measurement change

   First stored check of bediscoverable.ai.

   The baseline every later reading is set against, on the method in force then.

2. 5 Aug 2026Measurement change

   Engines stopped seeing the brand name while answering (brand-blind prompts, methodology v2).

   Naming the brand in the prompt inflated mentions; readings before and after are different quantities.

3. 8 Aug 2026Measurement change

   A source counts only when the answer itself attributed it, not when the engine merely retrieved it.

   Retrieved is not cited; the cited count is the honest one and is what the product reports since.

4. 9 Aug 2026Measurement change

   Six consumer surfaces read from logged-out sessions at a UK location (ChatGPT, Perplexity, Gemini, Copilot, Google AI Overviews, Google AI Mode) joined Claude.

   What buyers see is the consumer answer, not an API answer.

5. 31 Aug 2026Measurement change

   Methodology v5: one weighted score across all seven engines, weights from published reach.

   One number a customer can follow, with the weights and their sources on the methodology page.

6. 31 Aug 2026Site change

   Homepage and sitewide redesign (34 page files restyled).

   Pages that state the answer in the first sentence are the ones we see quoted in AI answers.

7. 5 Sept 2026Site change

   Two earlier research claims corrected: a page claiming an analysis of 2 million AI answers, with no dataset behind it, was rewritten without the figure; the July accuracy study is now labelled as one assistant through its API everywhere it is quoted.

   A measurement company cannot carry a claim it cannot show the data for.

8. 6 Sept 2026Site change

   Engine coverage copy on every page now derives from the engine registry instead of typed lists; a build gate stops a typed list from shipping.

   About forty sentences still said paid plans cover three engines when the product had measured seven surfaces for four weeks.

9. 6 Sept 2026Site change

   Entity links fixed: the LinkedIn and YouTube profiles in our structured data belonged to other organisations and were removed; the alternate name Discoverable AI added.

   We had been telling search engines we were a Kent marketing agency.

10. 6 Sept 2026Site change

    IndexNow wired to the deploy so Bing hears about new and changed pages the same day; first-touch attribution added so we can see which page a customer arrived on.

    IndexNow pings were a manual checklist step with no record of ever running; they now run on every deploy with a log.

11. 6 Sept 2026Site change

    The Gemini tracker page offers three real Gemini answers about a visitor's category, as a sample; the ChatGPT and Gemini tracker pages show aggregate figures from our own stored checks; a duplicate guide was folded into its sibling and redirected.

    A page that promises Gemini tracking used to hand visitors a Claude demo.

What moved

Nothing can be said yet: movement needs two readings on one method, and the current method has none. The next weekly check is on Monday 7 September. This paragraph is written by the readings, not by us.