Help · 66 answers
What GLIMPSE by Weevo measures, with what margin of error, where the sources come from and what it costs: answers with numbers, limits included.
The product · 7 questions
The product
GLIMPSE by Weevo measures how often AI models mention a brand: it asks Gemini, ChatGPT or Claude the same question dozens or hundreds of times and states the margin. It records the brands mentioned in each answer and returns the fraction of answers in which each brand appears, with the confidence interval that fraction carries. It does not survey consumers and it is not a poll: it measures the behavior of the models.
The product
The percentage next to a brand in GLIMPSE by Weevo is the mention share: the percentage of valid answers in which the brand is mentioned at least once. An answer that mentions the brand five times counts only once, and the denominator is the answers that produced a readable result, not the answers requested. The mention share does not measure citations, links or position in the text: those are different metrics that other tools call by the same name, which is why two tools measuring the same brand give different numbers that are both correct.
The product
GLIMPSE is the acronym for Generative Language Insights for Market Positioning & Sentiment Evaluation: readings of generative language models used to evaluate the market positioning and the sentiment of brands. "by Weevo" names the developer, Weevo Srl Società Benefit. The full product name is GLIMPSE by Weevo.
The product
Because "AI visibility" is not a single thing. A brand can be mentioned without being cited as a source, cited without being mentioned, and recommended without either. A brand with a 90% mention share and a 0% citation share has a specific problem — the model mentions it, but when the model looks for a source it goes to someone else — which is different from the problem of a brand that is not mentioned at all. GLIMPSE by Weevo declares which of the three it is measuring inside every result, exported data included, so anyone comparing two tools knows whether they are comparing the same thing.
The product
A GLIMPSE by Weevo search returns the table of mentioned brands, each with its mention share and confidence interval; the terms the model uses to describe each brand, with their sentiment; the charts; the sources the model consulted, with the cited pages of each domain; the phrases the model typed to search the web; and the raw data. The table and the data can be downloaded as CSV and JSON.
The product
GLIMPSE by Weevo is for anyone who has to decide something based on how AI models talk about a market: a company that wants to know whether it is mentioned when someone asks for advice in its sector, an agency that has to show a client where the client stands against competitors, and content writers who want to know which pages the models actually read on a topic.
The product
Anyone can ask ChatGPT the question. What a single user cannot do is ask it two hundred times under identical conditions, count the results against a declared denominator and know how wide the interval around the resulting number is. A single answer does not say whether a brand always appears or appears once in ten answers, and the difference between the two is everything. In addition, a personal conversation has a memory and a context that influence the answer; GLIMPSE by Weevo starts clean every time.
Reliability · 6 questions
Reliability
The interval next to each percentage in GLIMPSE by Weevo is the 95% confidence interval, calculated with the Wilson method. It says where the share would fall if the same measurement were repeated: in 95% of cases it would stay inside that interval. With 100 answers and a measured share of 30%, the interval runs from about 22% to 40%. The more answers are collected, the narrower the interval gets — and without the interval a percentage alone does not say how much it can be relied on.
Reliability
The number of answers needed depends on how high the measured share is. On a share of 50%, which is the most expensive case, the margin is about ±20 points with 20 answers, ±9.6 with 100, ±4.4 with 500 and ±2.0 with 2,500. A GLIMPSE by Weevo search can be run with 20, 50, 100, 250, 500, 1,000 or 2,500 answers, and GLIMPSE by Weevo shows the margin next to each depth before the choice is made, so the choice is a calculation and not a hunch.
Reliability
Because what matters is not the margin in points but how large the margin is in proportion to the brand's share. With 200 answers the margin is about ±6.3 points on a 30% share and ±2.1 points on a 2% share: in points the second looks more precise, but relative to the value they are ±21% and ±107%. In the second case the interval is wider than the value itself, which means nothing is known. In a crowded market a brand's share is structurally small, and that is why the same depth buys much less precision.
Reliability
Requested answers are the answers that were purchased; valid answers are the ones that produced a readable result. A request can fail, time out or come back in a format that cannot be interpreted. Percentages are calculated on valid answers, and GLIMPSE by Weevo always shows all three numbers — requested, valid, not started — because lumping them into a single "failed" count would hide the difference between a model that returns the wrong format and a batch interrupted halfway.
Reliability
No, and it should not. Generative models are not deterministic: the same question produces different answers, and that variation is exactly the phenomenon GLIMPSE by Weevo measures. Two measurements of the same question will differ by an amount compatible with the confidence interval. If they differ by more, something real has happened — or the method has changed, and in that case GLIMPSE by Weevo flags it.
Reliability
GLIMPSE by Weevo states it next to the difference. In comparisons and scheduled searches, every difference between two measurements of the same brand has its own 95% interval, calculated on the answer counts with Newcombe's method, which combines the Wilson intervals of the two shares: if the interval excludes zero the difference is real, otherwise it is noise, meaning it falls within the variation the same question shows on its own from one measurement to the next. With 100 answers per measurement, going from 30% to 45% is a real difference, with an interval from +1.6 to +27.7 points; with 20 answers, going from 25% to 45% is noise, with an interval from −9 to +45 points. Only real differences are colored green or red. In the comparisons saved in GLIMPSE by Weevo in October 2026, 706 differences out of 1,255 turned out to be noise.
The questions · 8 questions
The questions
A good question for GLIMPSE by Weevo is written the way a person with a problem, who does not yet know the solution, would write it. "What are the best yogurts for breakfast" works; "tell me about Danone" does not, because it already contains the answer. A question that names your brand measures whether the model can describe the brand, not whether the model chooses it — two different things, useful at different stages.
The questions
Yes, with the guided setup. GLIMPSE by Weevo reads the homepage of your site, recognizes the brand and the market and proposes five questions, one per category. If the site cannot be read by software — which happens with sites that block bots or are built entirely in JavaScript — a written description of what the company does and where it operates is enough. The proposed questions can be edited, tested with a free taste and then measured all together.
The questions
Because a person does not reach a brand from a single direction. The five categories are: category ("best protein yogurts"), problem ("I can't digest lactose, what can I eat for breakfast"), comparison ("is Greek yogurt or skyr better"), recommendation ("which yogurt do you recommend if I work out") and brand (the only one in which your name can appear). A brand can dominate category questions and be absent from problem questions, which is where decisions are made. GLIMPSE by Weevo flags which categories are missing from your panel.
The questions
The Country field tells the model which market the question is about. "Best online accountants" has different answers in Italy and in Spain, and a question without a country leaves the choice to the model. The country is part of the method: if it changes between two runs of the same question, GLIMPSE by Weevo flags that the two points of the time series are not comparable.
The questions
GLIMPSE by Weevo queries Gemini, ChatGPT or Claude, one model per search, at the same cost in credits. The best starting point is the model your customers use. To find out whether the models see a market the same way, the same question is put to each model in three separate searches: on the same question two models can mention completely different brands, and the difference is itself information.
The questions
No, and if it happened GLIMPSE by Weevo would say so. Every search declares the models that actually wrote the answers and how many each one wrote, for example "gpt-4.1-mini-2025-04-14 (100)". When a model is overloaded or returns a temporary error, GLIMPSE by Weevo retries the same model; if the error persists, that answer is not counted and its credit is refunded. Another model is used only if the chosen one no longer exists, and in that case the search is flagged as "Several models": a measurement that mixes two models is not comparable with one made by a single model. The method log compares the declared models between one run of a time series and the next, including new versions of the same model.
The questions
The right depth depends on what the number is for. To find out whether a market is worth measuring, a trial from the guided setup is enough: ten answers from Gemini and ten from ChatGPT, twenty in all. The interval is very wide, but the brands present show up immediately. Full analyses start at 20 answers, the Exploratory depth, and a number to put in a document needs at least 100. To follow a brand over time, the best approach is to choose one depth and keep it: changing the depth does not make the points non-comparable, but it changes the width of the interval from one point to the next.
The questions
With web search on, the model consults real pages before answering, and GLIMPSE by Weevo can record which pages. This applies to all three models. Without web search, the model answers from what it holds in memory: the answers remain valid for the mention share, but there are no sources to analyze and the citation share is zero by construction, not by measurement. To find out who cites a brand, web search has to be on.
Sources · 8 questions
Sources
The citation share in GLIMPSE by Weevo is the percentage of valid answers in which one of the domains you declared as yours appears among the sources the model consulted. It has the same denominator as the mention share, so the two numbers sit on the same page and are read together. They are not added together and one does not replace the other: they measure two different events.
Sources
All the domains you control, not just the main site. A product microsite, a store listing or a company page on a social network are your own channels, and in the answers they are indistinguishable from third-party sources: a measurement of the main domain alone reads zero while the model is citing the brand through the microsite. When several channels are declared, GLIMPSE by Weevo shows the share of each one and uses the share of the most-cited channel as the overall share — not the sum, because two channels cited in the same answer would count the same answer twice.
Sources
The Brand center is the GLIMPSE by Weevo page where every declared site has its own card: the questions that concern the site with mention share and citation share, the traffic that came from the assistants, who backs the site as a source and what the site is missing. Questions are linked to the site by hand, because only someone who knows the market can tell whether a question in which the brand does not appear belongs to another sector or to the brand's own; GLIMPSE by Weevo proposes the questions in which the brand appears or the site is cited, and also those that ask the same thing as a search in which the brand does appear: a question in your market where you are never named says as much about you as one where you are. At the bottom of the card there are suggestions, each one derived from a data point about that site, and if the data point does not exist the suggestion does not appear.
Sources
Yes. For every question linked to a site, GLIMPSE by Weevo shows how many sources back the brand and how many of those are the brand's own pages versus other sites' pages. In the list of sources, a brand's own site is marked as such, and a third-party source that mentions only one brand is marked as exclusive. Being cited only by your own site is a fragile position and a different one from not being cited at all: it means the model finds the brand but nobody else confirms it.
Sources
Yes. For every domain among the sources, GLIMPSE by Weevo lists the individual cited pages and in how many answers each one appears, not just the domain name. A sector ranking and a support page on the same site are two different levers, and knowing which of the two the model reads shows where to act. Gemini does not return the page address but a temporary redirect link that expires after a few weeks: GLIMPSE by Weevo resolves it as soon as the answer arrives, so the page is kept even after the redirect link stops working.
Sources
The "pages that back your competitors and not you" are the third-party domains that the model uses as sources to back other brands and never for your brand. The list is not a metric: it is a list of places where the brand is absent and that can be worked on. The competitors' own sites are left out, because they are not pages on which your brand could appear.
Sources
Yes, and GLIMPSE by Weevo flags it. It happens when a page on your site lists other brands — a guide, a comparison, a list of suppliers — and the model uses that page as a source to mention those brands and not yours. It is visibility given away: the work is yours and the benefit goes to others, and revising the page is usually enough.
Sources
With Gemini and with Claude, yes. When web search is on, both models return the searches they actually typed, and GLIMPSE by Weevo counts them: if the same phrase appears in 39 answers out of 50, that phrase is the door through which the model enters the topic. It is the most direct evidence available, because it shows not which pages mention a brand but what is typed to find them — and a page can be written for those words. From each phrase a new search can be launched with one click, to measure that phrase too. With ChatGPT this data is not available: its API returns the sources but not the searches, and GLIMPSE by Weevo does not invent them.
Sentiment · 3 questions
Sentiment
In every answer GLIMPSE by Weevo asks the model for up to five terms describing each mentioned brand — "creamy", "expensive", "suitable for children" — and counts them across all answers. The result is the list of terms associated with a brand and how many times each occurs, which shows how the model positions the brand against the others: two brands with the same mention share can be described in opposite ways.
Sentiment
The sentiment in GLIMPSE by Weevo is the judgment expressed by the terms associated with a brand. Each term is classified as positive, neutral or negative with a score from −1 to +1, and the brand's sentiment is the average of its terms' scores: above +0.05 it is positive, below −0.05 negative, in between neutral. GLIMPSE by Weevo also shows the distribution of terms among positive, neutral and negative, and the trend over time when the same question is repeated.
Sentiment
No. Shares, intervals and sources are calculated first and independently of the sentiment. If the classification of the terms fails, the search remains valid with all its numbers and the sentiment is reported as missing, not neutral: a score of zero is a measurement, and the absence of a score is something else.
Traffic · 7 questions
Traffic
Measuring the answers alone cannot settle it: whether appearing in AI answers brings visits is shown by the site's logs. By connecting a Google Analytics property, GLIMPSE by Weevo shows, next to the mention share, how many sessions came from each assistant and what those sessions did: a brand mentioned everywhere that brings one percent of the traffic and an absent brand that brings twenty percent tell two different stories. For sites hosted on Weevo's servers, measurements read directly from the server logs are added.
Traffic
Connecting Google Analytics to GLIMPSE by Weevo requires the GA4 property number — the one Analytics shows under Admin, not the G-XXXXXXX code placed on the pages — and read access granted to a service account, whose address GLIMPSE by Weevo shows on the page itself. Access is granted only once, from the property's user management. From then on the data is refreshed automatically every day.
Traffic
Server measurements are the counts read from the logs of the servers hosting the site: how many pages AI crawlers read and how many visits arrived by clicking from an assistant. Unlike Analytics, they also count visitors who do not accept cookies and visitors who do not run JavaScript. They are available for sites hosted on Weevo's servers, after Weevo has confirmed that the domain belongs to whoever declares it, and they are updated daily, with figures for each day and each page.
Traffic
The Brand center shows one row per assistant with three numbers: in how many of your questions that assistant mentions your brand, how many clicks from that assistant the server logs recorded and how many sessions Analytics recorded. Columns that cannot be filled are hidden instead of showing zeros. An assistant that sends visits but that GLIMPSE by Weevo does not query, such as Perplexity or Copilot, appears with its clicks and without the mentions column. When both server logs and Analytics exist, the Brand center compares them over the same days and states what share of the visits from assistants Analytics actually captures.
Traffic
Because Analytics only sees visitors who accepted cookies and have JavaScript enabled: it always undercounts, by an amount nobody knows. On a news site measured both ways over the same days, in September 2026, Analytics recorded 67 sessions from assistants against 92 visits counted by the server logs: it saw about three out of four. For this reason GLIMPSE by Weevo never adds the two numbers together and never puts them in the same time series: where two independent measurements of the same phenomenon exist, the gap between them is itself data — and that is how broken counts are discovered. For the same reason, the engagement of visitors arriving from assistants should be compared with that of visitors arriving from Google search, not with the site average: on the same site three sessions out of four arrived as "direct" with 5% engagement, probably automated traffic, and the site average was 16%, against 46% for visitors from assistants and 53% for visitors from Google. GLIMPSE by Weevo shows the three numbers together.
Traffic
When ChatGPT, Claude or Perplexity search the web to answer someone, they fetch the page at that moment with a dedicated agent: that is an on-demand read, and it means the page was used to write an answer for someone, whether or not that person then clicked. For sites hosted on Weevo's servers, GLIMPSE by Weevo counts reads day by day and page by page, only from the addresses published by the operators, and sets them next to the visits: the ratio of visits to reads is the landing rate. On a news site, over 29 days, ChatGPT read the site's pages 10,554 times while answering, and 57 visits followed: a landing rate of 0.54%. The landing rate is an upper bound: ChatGPT can also cite a page from its own index without fetching it, and Gemini, AI Overviews and Copilot leave no reads. Claude's reads cannot be verified, because Anthropic does not publish its addresses, and GLIMPSE by Weevo declares them as such.
Traffic
For sites hosted on Weevo's servers, the GLIMPSE by Weevo Brand center lists the pages that people reached by clicking in an assistant's answer, with the assistant they came from, how many times that page was read to answer and in how many of your searches GLIMPSE by Weevo saw it cited. A page that brings visits and that none of your searches cites answers a question you are not measuring: it is the most direct way to choose which questions to add. Next to them are the pages most often read to answer but never visited: assistants use them, yet nobody who reads the answer goes on to the site. On a news site, in October 2026, of the 67 pages visited from assistants only 3 appeared among the sources of the 21 searches made about that site.
Over time · 5 questions
Over time
Yes, weekly, monthly, quarterly or every six months. Each run of a scheduled search in GLIMPSE by Weevo is a complete search and uses credits as such, and it keeps the question text and the options it was run with. The runs table states whether each brand's latest change is real or noise. A question measured only once does not make a time series: at least three points are needed to see a trend.
Over time
No. A GLIMPSE by Weevo search with many answers can take several minutes and continues even with the page closed. From the search page, an email notification can be requested for when the search is complete.
Over time
GLIMPSE by Weevo says so. If the question text, the provider, the model (including a new version of the same model), the country, the web search setting or the metric changed between two runs, the time series is flagged as non-homogeneous and the method log shows where and when it changed. The log also records changes that are not part of the individual search, such as reapplying name corrections or a change in how the sentiment is calculated. This is why many "collapses" in visibility data are really changes of instrument: a step in the chart between two different measurements is not news about the market.
Over time
No. Changing the depth does not change what is measured, it changes how narrow the interval is: GLIMPSE by Weevo notes the change in the method log but does not flag the time series as non-comparable. Changing the question, the provider, the model or the country does.
Over time
Yes, from two to four at a time, among your own searches and those shared with you. A GLIMPSE by Weevo comparison sets the brands' mention shares and sentiment side by side. When the searches are runs of the same question, the comparison reads as a time series and shows the method log; when they are different questions, the method log does not appear, because there is no time series to protect. Next to the difference between the last and the first search, GLIMPSE by Weevo states whether it is real or noise. A comparison can be saved as a report, with its own public link and its own CSV.
Names · 2 questions
Names
GLIMPSE by Weevo automatically merges spellings that reduce to the same entity — "Danone" and "Danone S.p.A." — and shows which merges it made, because some will be wrong. No automatic rule can know that "Armani" and "Armani Exchange" are two brands while "Coca-Cola" and "Coke" are one: only someone who knows the market knows that.
Names
Yes, and the correction applies to your own searches. In GLIMPSE by Weevo you can merge two names that were kept separate or split two names that were merged, and reapply the rules to past searches without running them again and without spending credits: the aggregate stored with each search is enough to recombine the names. The reapplication is logged, because it rewrites numbers already produced, and whoever looks at that chart a month later needs to know that the step is not the market. A correction that holds for everyone — two genuinely distinct brands, not an opinion about the market — can be made general by Weevo after verification.
Reports · 7 questions
Reports
A GLIMPSE by Weevo report is a reading of a completed search designed for people who do not use the tool: the infographic of the top brands, the results table, the charts of mentions, sentiment, share versus sentiment and term distribution, and a map of the terms associated with each brand. Next to the table and charts, an artificial intelligence writes an analysis in words. A report is generated from a completed search or from a comparison of searches, and in both cases its data can be downloaded as CSV.
Reports
A GLIMPSE by Weevo report generated from a comparison gathers on a single page the table with one column per search (mention share and sentiment for every brand), the charts of the same measures, the terms the models use to describe the brands, the sources cited, the phrases the models typed to search the web and, when the searches are runs of the same question, the method log. An artificial intelligence writes three analyses in words: one on the table, one on the mentions chart and one on the sentiment chart, and they can be corrected before publishing. A comparison report can be published with a public link and downloaded as CSV.
Reports
Phil is the assistant inside GLIMPSE by Weevo reports. For each brand in the table Phil writes a comment with strengths, areas for improvement and advice, based on the data of that search, and answers questions in a conversation that can be downloaded. Phil interprets the numbers but does not produce them: shares, intervals and sources remain the ones calculated by GLIMPSE by Weevo, and Phil writes in the language of the interface. The brand-by-brand comments belong to reports generated from a search; a comparison report carries the analyses next to the table and charts, not Phil.
Reports
Yes, a GLIMPSE by Weevo search can be shared with another registered user by entering that user's email. Reports, on the other hand, can be published with a public link, readable without an account: that is the right form for a client who does not need to log in to the tool. A public report can be made private again at any time.
Reports
Yes. Every search can be downloaded as CSV, for a spreadsheet, and as JSON, with the complete results for processing by software. Report data can also be downloaded as CSV: the file of a report generated from a comparison has one column per search and, next to them, the difference between the last and the first. Exported files declare the measured metric, so a number copied elsewhere carries its meaning with it.
Reports
GLIMPSE by Weevo organizes searches with project folders: a search can be placed in one or more folders, by client, by market or by period, and the list of searches can be filtered by folder.
Reports
Yes, for partners who request it from Weevo. With the GLIMPSE by Weevo API a program follows the same path as the guided setup: it analyzes a site and receives the proposed questions, launches one or more questions on Gemini, ChatGPT or Claude at the chosen depth, from 10 to 2,500 answers, and receives the results (shares, intervals and sources), optionally with a signed notification when a search is complete. It costs the same as the site, one credit per answer, and every search states how many credits it used and how many were refunded. Keys are issued by Weevo, one per partner, and can be revoked at any time.
Costs · 3 questions
Costs
A GLIMPSE by Weevo search costs one credit per answer, with any of the three models: a search with 100 answers costs 100 credits. Credits are sold in packs of 50, 100, 200, 500, 1,000 and 10,000, always at €0.40 each — there is no volume discount, so buying more than needed does not save money. There is no subscription and credits do not expire. Every purchase and every use of credits appears on the credit movements page.
Costs
Yes. The GLIMPSE by Weevo guided setup analyzes your site, proposes the questions and measures one of them on Gemini and on ChatGPT with ten answers each, without using credits. There are three free trials; after that, the trial can still be run for twenty credits each time, with a notice before confirming. A free taste of ten answers has an interval about twice as wide as that of an analysis of fifty answers: it is for checking whether the question is right, not for putting its number in a document.
Costs
GLIMPSE by Weevo deducts credits at launch, based on the chosen depth. If the batch is interrupted, it is closed and the credits for the answers not produced are refunded. A search that has been stalled for too long is closed automatically: it does not stay "in progress" forever.
Account · 3 questions
Account
The GLIMPSE by Weevo interface is available in eight languages: Italian, English, Spanish, French, German, Polish, Brazilian Portuguese and Uzbek. New users start in their browser's language and can change it in their profile: that language is used for the interface, for the account emails — the address confirmation and the password reset — and for the texts written by the artificial intelligence, Phil included. Questions can be asked in any language, and it is best to ask them in the language of the market being measured: the answers change.
Account
With two-step verification: in addition to the password, at login GLIMPSE by Weevo asks for a code generated by an authenticator app on the phone. When two-step verification is turned on, GLIMPSE by Weevo also provides recovery codes to keep safe, which are needed if the phone is not available. Turning it on and off requires confirming the password.
Account
Searches stay in your GLIMPSE by Weevo account and are visible only to you, to the people you share them with and to anyone who opens a report you published. Deleting the account is permanent and erases all searches. To the artificial intelligence providers — Google, OpenAI and Anthropic — GLIMPSE by Weevo sends only what each function needs: the question text for searches; the homepage text or the description you write when the guided setup proposes questions; the terms associated with brands to calculate the sentiment; the results of a search when a report or Phil writes its analysis. Your account data, such as name and email, is not sent.
Limits · 5 questions
Limits
GLIMPSE by Weevo cannot know what the model will answer to the real person who asks that question tomorrow, because models change and answers are not deterministic. It measures a distribution over a sample, at one point in time, with a declared method — which is more than a single answer says, and less than a certainty.
Limits
GLIMPSE by Weevo queries Gemini, ChatGPT and Claude. Perplexity, Copilot and the others are not queried: if the traffic to your site shows visits coming from them, GLIMPSE by Weevo reports it but cannot explain it. The asymmetry is declared, not hidden.
Limits
With caution. With web search on, Gemini almost always links the answer to the pages it consulted, while ChatGPT indicates sources within the text in only part of its answers: about one in three in the tests GLIMPSE by Weevo ran in September 2026. ChatGPT's citation share therefore comes out lower because of the way the model answers, not necessarily because brands are cited less. Citation shares are compared over time on the same model; across different models they mostly show how each model uses sources.
Limits
Yes, and they are documented. A domain that replaces a symbol in the brand name with a digit — a brand "A&B" with the site a3b.it — is not recognized as that brand's site, and ends up among third-party pages. GLIMPSE by Weevo checks automatically every day that these cases do not produce absurd claims, but name recognition remains imperfect and it is worth knowing that.
Limits
The sources in GLIMPSE by Weevo are the ones the model declares it consulted, and that is a fact. Attributing a source to a specific brand, instead, is an inference: when a passage of the answer mentions several brands, GLIMPSE by Weevo attributes that source to none of them, because there is no way to know which of the brands the source backs. Some correct attributions are lost, but no false ones are invented.
Support · 2 questions
Support
To get support for GLIMPSE by Weevo, write to [email protected]. If the question concerns a specific search, the search number — the one shown next to the title — makes it possible to answer by looking at the real data instead of guessing.
Support
Report it, with the search number. Several defects in GLIMPSE by Weevo were found by users who looked at a result and said "but that's not true": a wrong attribution, a name merge that joined two different companies, a margin that hid how large it was relative to the share. Each time, the fix went into the product and into a test that prevents the defect from coming back.
Write to [email protected]. If your question is about a search, include its number: we answer by looking at the real data instead of guessing.
Published on updated on