We measure the assistants buyers actually use
Buyers don’t ask the benchmark champion. They ask the assistant they already have open, in its default mode. So Kaivox probes the consumer-default tier of each assistant, not the heavyweight models a buyer has to go looking for. We measure five assistants: ChatGPT, Gemini, Claude, Google AI Mode and Perplexity.
openai/gpt-5.6-lunaThe model ChatGPT gives a free user, at the quick Instant setting free users get (OpenAI's models page, October 25, 2026). OpenAI is rolling a newer model out to paying users first; when free users get it, we follow.
google/gemini-3.5-flash-liteThe model the Gemini app gives a free user for most questions, set to think only briefly before it answers, as a free user gets it (Google's models page, October 27, 2026). When Gemini searches the web before answering, that answer comes from the same model, Gemini 3.5 Flash-Lite.
anthropic/claude-sonnet-5.5The model the Claude app gives free and Pro users, at the Medium setting the apps use (Anthropic, October 2026).
google/ai-modeGoogle's AI chat, read from the real Google page through a data provider. There is no model or setting for us to pick: it answers as Google serves it, from your town if you are a local business, from your state if you are a regional one, otherwise from the United States. It always searches. Google does not publish how much buyers use it beside the other assistants, so its share of your score, about a tenth, is our estimate, built from two published sources: SparkToro's finding that 0.34 percent of US Google searches moved into AI Mode (January to April 2026, from Similarweb's clickstream panel, with Google's phone app left out), and Similarweb's monthly visits to google.com (September 2026) and to the other assistants' sites. Leaving out the phone app likely makes it read low. It applies a share of searches to a count of visits, so it is an estimate, not a count.
perplexity/sonarFree Perplexity picks a model per question and does not say which; Sonar is the closest model Perplexity offers to businesses. It searches the web by nature, so it appears only in the live layer below.
ChatGPT, Gemini, Claude and Perplexity have been the measured panel since June 1, 2026, and Google AI Mode joined on November 3, 2026. The model behind each of the first four, and how long it thinks before answering, follows what the app serves a free user, most recently re-aimed on October 27, 2026. Every scan records exactly which models answered it, and any change to the lineup or to the scoring ships as a new measurement version, listed in section 09 below with its date, so a moving trend line can always be traced to either the market or the instrument.
read on your Google searchesThe AI answer Google writes at the top of its results. It shows up on Google searches, not on questions typed into a chat app, so we read it on every Google search we track for you, three times per scan: twice as a desktop search and once as a phone search, from the United States for a business that sells nationwide or from its state for a local one. We look more than once because the answer changes from one look to the next. We count how many of the answers Google gave said your name or linked to your website, and who they named. Searches for your own name are counted apart. It has its own count and does not change your AI visibility score, because people type different things into Google than into a chat app. Google AI Mode is different: it is a chat, so it gets the buyer questions and counts in the score.
Two layers: what AI remembers, what AI finds
Three of the five assistants are measured two ways, because an AI answer has two sources and they move at different speeds. Perplexity and Google AI Mode always search, so they appear only in the live layer.
Memory
What the AI remembers about you
The model answers from its training data alone. This layer reflects your long-term footprint on the web and moves slowly, when models retrain. It is the answer buyers get when the assistant doesn’t bother to search.
Live
What the AI finds when it looks
The same assistants with live web search (Gemini’s live answers come from Google’s own search), which is how the consumer apps increasingly answer buying questions. This layer moves when content gets published, in weeks rather than quarters, and it is where a fix shows up first.
A gap between your memory score and your live score is a finding, not noise: strong memory with a weak live score means your fresh content is invisible to retrieval; strong live with weak memory means AI finds you but doesn’t yet know you. The two diagnoses have different fixes, which is the point of measuring both.
The questions are real buying questions
Each scan asks the assistants what a real buyer in your category would actually type, and the questions are shaped to your business. A national brand is measured on discovery and comparison questions like “best [category] for [need]” or “[rival A] vs [rival B]”. A local business is measured the way people really search for one: “who should I hire for [service] near me,” not a national best-of list. Across both, the set spans discovery, head-to-head comparisons against the rivals that truly compete with you, alternatives, problem-first questions, and trust checks.
We don’t invent these questions in a vacuum. They are grounded in real demand in your category: the keywords people actually search and the “People Also Ask” questions Google surfaces, so the panel reflects how buyers really phrase things rather than how we imagine they do. Two hard rules hold throughout: the questions behind your score never name your brand, because the entire point is whether the AI brings you up on its own, and they are grounded in today’s date, so a search-backed assistant is never steered into last year’s articles. We also ask a few questions that do name you, to read what the AIs say about you, and we report those apart from the score.
You see the full question set, the exact wording of every question, and approve it before any scan runs, and every scan asks that same approved set. That is what makes your score comparable scan over scan: when the number moves, it moves because the assistants’ answers changed, not because we asked different questions. And when the set does change, because you edited it, the app says so on the score and stops comparing across the change rather than passing the difference off as movement. (The free check picks its own questions and shows them on the result; the approval step belongs to the full scan.)
Asked up to five times, because once is a coin flip
The same question asked twice can return two different brand lists. A single-sample score cannot tell a real change from measurement jitter, so every question is asked up to five times per assistant (we stop early only after at least two asks, and only when every answer so far agrees on whether you were named and how prominently; a question counts the same in your score however many times we asked it), and your score is reported with a repeatability band, for example 27.8 ± 2.0. That band is calibrated the hard way: we periodically re-scan brands back to back with nothing changed and measure how far the number moves on its own (the current floor comes from three back-to-back scans of one unchanged brand; same-day repeats on three more brands came in tighter, so the published band is the conservative one), and the band you see is never allowed to claim more precision than those controlled repeats actually showed. When a scan’s own spread comes in narrower than that calibrated floor, we show the floor and mark it “calibrated minimum band”, which is why two different brands can carry the same band. The band is one scan’s wobble. The gap between two scans wobbles more, about 1.4 times the band, and we call a move real at the usual 95 percent line, 1.96 times that. At the 2.0 minimum band that comes to 5.5, which we round up to 6: a move of 6 points or more on the 0-100 score counts as real. A scan with a wider band prints a higher bar beside its score, by the same arithmetic. The app says the same thing in place: a small score move is labeled “within normal variation” right beside the number. Single-sample measurement dressed up as precision is the most common shortcut in this category; sampling every question, calibrating against real repeat runs, and showing you the band is how you can tell we’re not taking it.
The size of the instrument, in plain numbers: a full scan asks about 40 questions, each of them up to five times on each of eight assistant-and-mode combinations (the five assistants: three asked from memory and again with live search, plus Perplexity and Google AI Mode, which always search). On a recent full scan that came to 38 questions and 568 graded answers. It would be cheaper to ask hundreds of questions once each. We spend the budget on the repeats and on the calibrated noise floor instead, because the second ask is what tells a real change from a wobble, and a trend you cannot trust is not worth having more of.
What the score means
The Kaivox AI Visibility Score (0 to 100) blends four measured signals: how often you appear at all, how prominently you appear when you do (leading an answer counts for more than a passing mention), your share of voice against the rivals the assistants actually name, and how often your own site gets cited as a source. Each assistant is scored separately, then blended. No signal in the score is assumed, padded, or simulated; if we didn’t measure it, it isn’t in the number.
The blend is not a flat average. Each assistant’s influence on your score reflects how much it is used on the web, from published traffic figures, dampened so no single assistant dominates the number and smaller assistants still register; ChatGPT carries the largest share because it carries the most use. Public usage figures disagree with each other, so the weights are our dampened reading of them, applied the same way to every brand. When we recalibrate these weights against published usage data, the change ships as a new measurement version (section 09), every scan carries its version stamp, and no trend line ever presents a recalibration as market movement.
The score is built only from questions where the assistant actually recommends a business. Pure how-to and cost questions, where the AI explains something and names no one, are measured separately as Answer Visibility, so they never drag your recommendation score down. And when an assistant declines to name anyone at all, that answer is set aside rather than counted as you being absent, because in some categories the assistant routinely refuses to name names. The number reflects the questions where being recommended was genuinely on the table.
Your corrections outrank our scan
Kaivox reads your brand from the public web, and the public web is sometimes wrong about you. Anything you correct, from your category to the names your brand goes by, becomes permanent ground truth: it overrides what we scanned, survives every rescan, and steers every future measurement. The system also audits its own data after every scan and shows you what looks wrong, so finding problems doesn’t depend on you stumbling into them.
How a fix is proved
A score tells you where you stand. The proof loop tells you whether something you published changed it. When Kaivox publishes a page or post against one buyer question you were losing, it records a baseline that day: the same question, asked twelve times of each assistant with live web search on. Fourteen days later it asks the same question the same way, twelve times again, and compares the two ends.
Two things stop a coincidence from being sold as a win. First, two untouched questions from the same scan, questions the content did not target (the ones whose answers name you least), are re-asked alongside it at both ends: if those moved too, the market moved, and the lift is not credited to your content. Second, a noise test: an exact permutation test at the five-percent level, which asks how often a gap this large would appear if the content had changed nothing. A move is credited only when it is positive, it clears that test, and the untouched questions stayed flat. If one of the two untouched questions cannot be measured at both ends, the check runs on the other. If neither can, no credit is given.
The verdicts are the ones the instrument can actually give: a real, attributable lift; a real lift that is small (under five points on that one question: it counts as a win, but on its own it is not enough to say “do more of this”); a move that the untouched questions also made; not enough evidence yet (12 re-asks at each end can confirm only a big move, so this is not proof the fix failed); and a real drop. A lift that is not proved at day 14 is checked again at day 28, and a proved lift is checked again 28 days after it was proved, to see whether it held. Every result stays on the board, the ones that did not prove out next to the ones that did.
Two limits, stated plainly. The number that moves is the score for that one question, not your overall visibility score. And the range shown beside a result is the spread of the same-day re-asks at each end, not a measure of how sure we are that your content caused the change; the untouched questions are what carry that weight.
What this measurement is not
- 01
It is not every conversation. AI answers vary by user, location, and chat history. We measure a representative, repeatable slice under controlled conditions, which is what makes scan-over-scan comparison meaningful.
- 02
It is not comparable across tools. Different products probe different models with different questions; a Kaivox score is built to be compared with your last Kaivox score and with competitors measured in the same scan, not with a number from another vendor.
- 03
It is not static. The assistants change under everyone constantly. Our lineup discipline, date grounding, and repeatability bands exist so that when your trend moves, you can trust the movement is yours.
- 04
It is not a margin of error. The plus-or-minus band is a repeatability band, not a statistical confidence interval: it is never narrower than the floor our controlled repeat runs demonstrated, and it widens when a scan’s own spread exceeds that floor. In a small or niche market, once off-topic questions and refusals are set aside, a score can rest on a modest number of answers, so treat a small move as noise and a sustained trend as signal.
- 05
It is not a cross-brand leaderboard. The 0-to-100 score is built to track ONE brand over time and against the rivals named in the same scan. Comparing the raw score of two unrelated brands, or two different categories, is not something the scale is built to support.
- 06
It is not immune to the models changing under their own names. We hold the assistant lineup fixed and compare like-for-like scans, but when a provider updates a model without renaming it, a score can shift for reasons unrelated to you, which is why a sustained trend matters more than any single scan-to-scan move. We check the lineup against what the free apps serve, and every change ships as a dated measurement update in section 09, so your trend is cut only when the instrument really changed.
- 07
It is not every kind of AI answer. We ask the assistants through their business interfaces, so paid placements, the ChatGPT shopping carousel and product picks drawn from merchant feeds are not seen, and Copilot, Meta AI and Grok are not in the lineup yet. Google AI Mode is the exception to the first sentence: we read it from the real Google page, and we leave out any paid placement in it. Google’s AI Overviews are read on your Google searches and reported beside the score, not inside it (section 01).
- 08
It is not pinned to your customer’s exact location. When an assistant searches the web live, that search runs from our infrastructure, not your buyer’s town. Google AI Mode is the exception: we ask it from your town if you are a local business, from your state if you are a regional one, otherwise from the United States (section 01). Local questions name the place so the answers stay locally relevant, but the search origin is not your customer’s address.
- 09
It is not a transcript of the consumer apps. We ask each assistant through its business interface, not by driving the consumer app, with a neutral setup, no chat history, no saved memories and no personal settings. Google AI Mode is the exception: we read it from the real Google page through a data provider, with no setup of ours. Published research shows the two routes differ. On the same product questions, ChatGPT’s app and its business interface shared about 12 percent of the websites they displayed, and Gemini’s about 15 percent (Uberti-Bona Marin and colleagues, September 2026, arXiv 2609.18729). On benchmark questions, the apps repeated their own answers about 2 percentage points less often (Wang and colleagues, September 2026, arXiv 2609.08861). So read a Kaivox score as a controlled, repeatable measure of change, not as what one buyer sees in the app: we ask the same way every scan and flag any change to that, so how we ask is not what moves your trend.
What changed, and when
Every scan is stamped with the version of the measurement that produced it. This is the current version and every change before it, in plain language. Your stored scans never change; only new scans use the new version, and when a trend line crosses one of these dates the app says so beside the number and does not read the step as market movement. These changes run both ways: some can only lower a score, others raise one, and each entry says what it does to your trend line.
CurrentMeasurement version 18since November 12, 2026
- Each assistant's share of your score follows newer usage figures, and a cut-down company name no longer counts as that companyv18 · November 12, 2026
Each assistant counts toward your score by how much it is used on the web, from published traffic figures. We even this out so the biggest one does not drown out the rest. Our figures had fallen behind: Claude is used more than they showed, so it now counts for more, and ChatGPT and Perplexity count for less. Separately, when a company's name ends in Inc., LLC or Co., we also look for it without that ending. When what was left was one plain word, like "Pools" from "Pools Inc.", it also matched other firms' names, such as "Blue Haven Pools", and ordinary sentences, and gave that company their mentions. A one-word name now counts only when it was recorded as that company's name. A shortened name that ends in Inc. or Co. also no longer matches the start of a longer word, such as "Morrow Co" inside "Morrow Corporation"; it still matches its own full word, such as "Morrow Company". The questions we ask did not change.
On your trend line: Treat scores from before this date as a separate series. The same answers can read a little higher or lower, depending on which assistants name you. A competitor whose name is one plain word plus Inc. or Co. may show fewer mentions from your next scan.
- Google AI Mode joins the assistants, and the score now reaches 100v17 · November 3, 2026
We now ask Google AI Mode the same buyer questions we ask the other four assistants, searched from your area if you are a local or regional business, and read its answers the same way. It counts for about a tenth of your score. No public count measures how much buyers use it beside the others, so that tenth is an estimate from the best figures published, and we will revisit it when a direct count exists. Separately, the score used to stop at 90 because its parts added up to 90. It now runs to a true 100, so the same answers read about 11 percent higher. One more correction: when we compare you with your top rival, an assistant citing the rival's site in its source list now counts for the rival, as it always counted for you. So a rival's score can rise by more than that 11 percent while yours does not.
On your trend line: Treat your scores from before this date as a separate series. A brand that stood still reads about 11 percent higher from the scale alone, then a little higher or lower from AI Mode, and a top rival can read a little higher again from the citation correction. None of these is a change in your market. Nothing before this date is re-scored.
- Gemini now answers with the model a free user getsv16 · October 27, 2026
Google says a free Gemini user gets its Flash-Lite model for most questions. It is set to think only briefly before it answers. We had been asking a bigger Gemini model. Free users get that one only if they pick it. We collect two kinds of Gemini answer: one from what the model already knows, and one where it looks things up on the web first. Both now come from Flash-Lite, set the way a free user gets it. When it looks things up, it lists about 7 web pages as sources in an answer. The old model listed about 11. We tried both models on two businesses' questions on the same day. Each business was named about as often either way, but a test that small cannot prove the two agree. ChatGPT, Claude and Perplexity are unchanged.
On your trend line: Treat your scores from before this date as a separate series. A change at this date can come from the new Gemini model, not from your business. Nothing before this date is re-scored.
- A different website with a similar address no longer counts as yoursv15 · October 27, 2026
An answer cites your site when it links to your website. We used to look for your web address anywhere in the answer. So a different company whose address contains yours (abcroofing.com contains bcroofing.com) counted as a link to you. So did your address with more on the end (bcroofing.com.au), and a review site's page about you when its link has your address in it (trustpilot.com/review/bcroofing.com). Now only links to your own website and its pages count. Rival websites are counted by the same rule. When we re-read the newest stored answers with the new rule, 14 of 6,152 changed.
On your trend line: If your citation count drops at this date, that is a correction: the old count included links that were not yours. Your visibility did not fall. Scores from before this date and after it are not compared, and earlier scores stay as they were measured.
- Two assistants now think as long as they do for a free userv14 · October 25, 2026
The models we ask did not change; how long two of them think before answering did. A free ChatGPT user gets its quick "Instant" setting, and we had been asking the same model at a setting that thinks first. The Claude app answers at its Medium setting, and we had been asking at the higher setting its business interface uses by default. Both now answer the way a buyer in the app sees it. Gemini and Perplexity are unchanged. ChatGPT is rolling out a newer model to paying users first; free users still get GPT-5.6 Luna, so that is still the model we ask.
On your trend line: Scores from before this date and after it are not comparable: two of the four assistants answer differently. Nothing before this date is re-scored, and scans on the new setting are compared only with each other.
- The lineup now asks the model each app gives a free user todayv13 · October 1, 2026
Our rule is to ask each assistant the way a buyer who just types a question gets it: the model the app serves by default, on the free tier where there is one, because that is most buyers. Three of the four had moved on. ChatGPT has given free users GPT-5.6 Luna since August 6 (paying Plus users get a different model); the Gemini app has given free users Gemini 3.6 Flash since July 21 (the newer Flash tiers are paid); and Claude has given both free and Pro users Claude Sonnet 5.5 since late September. Each of those is now the model we ask. Gemini's live-search answers still come through Google's own search, as before; the change there is in the memory answers. Perplexity is unchanged. Google AI Mode is still not in the lineup.
On your trend line: Scores from before this date and after it are not comparable: a different model answered most of the questions. A move at this date is the lineup catching up with what buyers see, not your market. Nothing before this date is re-scored, and scans on the new lineup are compared only with each other.
- An answer that only repeats the rival a question asked about no longer counts against youv12 · September 8, 2026
When a question names another brand, the assistants name that brand back almost every time. An answer that named only the brands the question handed it is now set aside and counted apart, because there was no room in it for anyone else; an answer that brought in any other name still counts, because that name could have been you. A rival named in a question no longer earns share of voice for being named back in the answers to it, and the competitor table says how often each rival was asked about by name. Rivals your questions ask about by name will show a smaller share than before; the brands the assistants volunteer on their own will show a larger one. The Claude assistant in the lineup now answers with the model the Claude app has used by default since June 30.
On your trend line: Scores from before this date can sit lower than the same brand measured after it, and share-of-voice numbers are not comparable across the date; the difference is the instrument, not the market. Scans on the new method are compared only with each other.
- A link to your site counts only when the page loadsv11 · August 28, 2026
Assistants sometimes invent page addresses on real sites. Each link to your own site in an answer is now checked once per scan, and a link to a page that does not exist (a confirmed 404) is no longer counted as a citation. A link that could not be checked, for example because the site blocked the check, still counts exactly as before, and the citation tile says when links were not checked.
On your trend line: This can only lower the citation part of a score, never raise it. Mentions, prominence and share of voice are untouched.
- A brand name must start where the match startsv10 · August 10, 2026
Your name is matched only at the start of a word. A different company whose name merely contains yours ("Pumpjack Golf" for "Jack Golf") no longer counts as you, and an ordinary word that happens to contain a short name no longer earns you a mention.
On your trend line: This can only remove mentions that were never yours. A handful of brands read slightly lower at this boundary and that is the truer number.
- Every repeat of a question counts equallyv9 · August 8, 2026
When a question is asked several times of one assistant, each real answer now carries equal weight. Before this, the order the answers happened to be stored in could give one answer four times the weight of another.
On your trend line: Movements at this boundary were under half a point on the headline score across our whole fleet. Repeat scans of an unchanged brand agree more closely from here on.
- A fix for brands whose answers name almost no rivalsv8 · July 24, 2026
When the assistants named almost no rival brand in your answers, the mention rate was being taken over the few answers that named someone, which could publish a score in the 70s for a brand named once in 36 answers. It is now taken over all the answers when rival signal is that thin.
On your trend line: Only scans with almost no rival signal were affected. Everything else is byte-identical.
- More of your real mentions and citations are countedv7 · July 8, 2026
Citations the assistant itself declared and that resolve to your website now count alongside links found in the answer text. Punctuation variants of your name (with or without periods) match. A numbered heading that names you carries its list position.
On your trend line: Scores can step up at this boundary because credit that was always real is now counted.
- Honest floorsv6 · July 5, 2026
A brand with no tracked rivals is scored on the same scale as everyone else. Share of voice is reported only when rival mentions were actually measured. A brand whose name is also an ordinary word is matched only in its exact cased form.
On your trend line: This can only lower a score, and only for the least-evidenced scans.
- Rivals judged by where they operate; local questions name the placev5 · July 4, 2026
A same-category business in the wrong state no longer counts as a rival in a local brand’s share of voice. New local question sets name the state. New national sets carry a few questions that name the brand, and those are kept out of the headline score.
On your trend line: Local brands may have seen wrong-state rivals drop out of their share of voice at this boundary.
- Memory and live answers blended by how often each answeredv4 · July 4, 2026
Within one assistant, the memory answer and the live-search answer are combined in proportion to how often each produced a usable answer, instead of half and half. A single borderline answer had been able to swing a score by several points between two identical scans.
On your trend line: Repeat scans became steadier. The answers underneath did not change.
- Assistants weighted by real-world usev3 · July 4, 2026
Each assistant’s share of your headline score now reflects how many people use it, dampened so no single assistant dominates and the smaller ones still register. Perplexity had been carrying far more weight than its reach.
On your trend line: The blend shifted at this boundary; the answers underneath did not change.
- Cleaner name matching and citation countingv2 · July 2, 2026
Your name is matched in its exact casing, so a common word no longer counts as your brand. A link that redirects to your site counts as a citation. An answer whose brand list could not be read is set aside instead of being counted as a miss.
On your trend line: Scores from before this date can sit a little differently from the same brand measured after it.
- The first versionv1 · before July 2, 2026
Every scan completed before July 2, 2026 is version 1.
Smaller changes to how we measure
These carry no version number because they do not change what a score means. Each note says what changed and for whom.
- Answers that did not search the web no longer count among the answers that could link your siteNovember 12, 2026
An assistant with web search on decides for itself whether to search. Until this date, an answer it gave without searching still counted when the Site tab and the Visibility page say how many web-search answers linked your site. Now those counts leave it out, and we record for every answer whether it searched.
On your trend line: Your score and its trend are unchanged. On the Site tab and the Visibility page, the number of web-search answers can be smaller from this date, by more for a business where Claude often answers without searching.
- We now also check the links in the assistants' source listsNovember 12, 2026
Each scan checks whether the links the assistants give for your site still load. Until this date the check read only the links written in an answer's text and skipped the list of sources an assistant shows beside it, where most links are. It now checks both, on every site that answers our check, and the same for your top competitor's links. Where your score counts links, a link that does not load is left out of it.
On your trend line: Replaying every brand's stored answers through the new check moves no score at one decimal. Two brands' link rates drop by under one percentage point, where an assistant kept linking a page that no longer exists.
- An answer that failed part way counts as no answerNovember 3, 2026
An answer an assistant's provider reports as having failed part way now counts as no answer, not as a full answer that never named you.
On your trend line: Stored answers cannot be checked for this, because the stop was never saved, so new scans may read a little differently.
- When an assistant cannot be reached, its answer is left out, never filled in by another modelOctober 25, 2026
Before this date, if a call to one of the four assistants failed, a backup model could answer in its place, and that answer was counted as the assistant's; Gemini's live answers could also come from a version of Gemini that had not searched Google. Now a failed call is recorded as not measured, and the scan asks again where it can. We checked every stored answer: 4 of 29,390 were a backup model's, all from one day (August 29), and none since.
On your trend line: Your score and its trend are unchanged; replaying stored answers moves no score. If an assistant cannot be reached for a whole scan, that scan is scored on the answers that came back, never on a stand-in.
- Questions that use your name are asked once per assistant, and we now read what the answers saySeptember 30, 2026
A question that already uses your name gets your name back in nearly every answer, so asking it up to five times told you nothing. Each assistant now answers it twice from memory and once with a web search, and we read those answers against your profile: a wrong fact, a mix-up with another business, a claim to check. Your score never counted these questions, and replaying every brand's latest scan changed no score.
On your trend line: Your score and its trend are unchanged. Two counts read fewer answers from that date, because the repeat answers to these questions are no longer asked: how the assistants talk about you, and on the Site tab how many answers linked your site (the part from questions that use your name).
- A name you tell us is not you no longer counts as youSeptember 19, 2026
When you name a business that is not yours in your aim consult (a company with a name close to yours), the assistants naming that business no longer count as naming you, and that business is measured as a competitor in its own right. Nothing changes for a brand that has not named one.
On your trend line: From that brand's next scan only. The consult that adds the name also refreshes the brand's question set, so its trend line restarts at the same time; no earlier score is changed.
The Google score: what changed, and when
Your Google score has its own version number. These are its changes. Stored scans keep the version they were scored under.
- A search for the person behind the business is a search for your namev10 · November 7, 2026
If your business is built around a named person (the owner, the doctor, the attorney), a Google search with that person's name in it, like "dr reid amedee chiropractor", now counts as a search for your name: reported on its own line and left out of the market score, the same as a search for the business name. Before this, such a search sat with the market searches because we only matched the business name and its other names, so ranking first for your own name read as market performance, on Google and on the Maps summary. We also now count your own listing in the map box as one of your results when we decide whether a name search is about you, since on a name search Google often shows you first in the map box and other businesses in the blue links.
On your trend line: A brand with a named person reads a little lower from this date, because a page-one win on that person's name has moved out of the market score; its Maps summary changes with it. On the brands we tested one moved, by 2 points, at its next scan; the rest did not change. Nothing before this date is re-scored.
- Google ranks now come from a different read, checked against a second onev9 · November 3, 2026
The Google results we had been buying came back wrong on most reads, and only some of the wrong pages could be spotted. We now read ranks from a different source and check page one against a second one. Where the two disagree on whether you are on page one, that search is marked not confirmed and counts neither way. The score now counts page one only: past page one a search still shows the rank we read, from one source, and it does not move the score. Searches answered by where you stand, like "chiropractor near me", moved to Maps for a local business, and a local business gets no Google score for now, because nothing we can buy reads Google from inside your town.
On your trend line: Do not read a line across this date. Most scores fall at this boundary because the part of the score that came from page two and deeper is gone, not because anything changed in your market.
- A search that asks for you by name stays a name searchv8 · September 26, 2026
A search with your own name in it, like "[your brand] promo code" or "[your brand] near me", used to count as a market search whenever Google filled its first page with other sites: businesses with similar names, coupon sites, or shops that sell your product. Someone typing it already knows you, so it now counts as a search for your name, reported on its own line and left out of the market score. Two things keep a search in the market: a comparison or alternatives search ("[your brand] vs [rival]", "brands like [your brand]"), and a brand whose name is made of the words its buyers type for the service, where Google shows other businesses because the words describe what you do.
On your trend line: A brand that ranked near the top on a search like this reads a little lower from this date, because a win on its own name has moved out of the market score. On the brands we tested, two brands moved, each down 3 to 4 points on its newest scan; the rest did not change. Nothing before this date is re-scored.
- We now check Google's first five pages, not tenv7 · September 26, 2026
For each tracked search we now ask our data source for Google's first five pages instead of the first ten. Since July, Google often stops showing results past the first few pages, so on a bad day about half of our deep checks came back incomplete. Five pages come back complete far more often. A site Google lists past page five now reads as not on the pages we checked; a rank past page five on an older scan was real and stays on that scan. When an answer stops before the five pages we asked for and your site is not in it, we ask once more and, if it is still short, count that search as unseen rather than not ranking, as before.
On your trend line: Two things can lower a score a little at this date with no change in your market. We no longer look past page five, so a site listed there now reads as not ranking on that search. And a search we could not fully see before, because Google's answer stopped early, now more often comes back complete and counts as checked. On the brands we tested the move was 0 to 3 points. Your average rank can read better at the same time as your ranking count falls, because a rank past page five no longer counts in the average, and a search where a rival is on page one and you sat past page five now counts as a gap. Nothing before this date is re-scored.
- A missing site on an answer that came back short is unseen, not unrankedv6 · September 22, 2026
For each tracked search we ask our data source for Google's first 100 results. About one answer in ten comes back shorter than that with no warning. When an answer holds fewer than 50 results and your site is not among them, we no longer count that as "not ranking": the row says how deep we could see, and we set that search aside from the part of the score that asks whether you rank at all until a fuller answer comes back, and if too many answers come back short the score is held back for that scan with the reason beside it. A search whose answer stops short on every scan for two weeks is treated as one Google itself answers briefly, and a missing site there counts as missing.
On your trend line: Scores can sit a little higher from this date, usually 1 to 3 points, and most for a brand whose site was missing from many answers that came back short; a scan with many cut-short answers shows "held back" instead of a number. The change is in how we count, not in your market.
- More spellings of your name count as your namev5 · September 17, 2026
A search typed with "and" in place of an "&", or without the "&", now counts as a search for your own name. And a search whose first page is all your own site and profiles is treated as a name search even when it is phrased like a category. Both keep name searches out of the market score and on their own line.
On your trend line: A brand with an "&" in its name, or one whose own site fills page one for a search phrased like a category, may read a little lower from this date, because a page-one win on its own name has moved out of the market score. Other brands are unchanged.
- A search for a rival by name is judged by what Google shows for itv4 · September 8, 2026
A search that names a rival counts with the market again when Google answers it with lists, comparisons and other rivals, because that is a page you can win. It stays out only when Google answers it with the rival's own pages.
On your trend line: A handful of rows moved back into the score at this boundary. Small changes either way.
- A search for a rival by name is not a search you can winv3 · September 4, 2026
A tracked search that is just a rival's name leaves the score and the gap count. A buyer typing it has already chosen.
On your trend line: A few brands read slightly higher or lower from this date; the searches underneath did not change.
- Searches for your own name are counted apartv2 · August 9, 2026
The score runs over searches for what you sell. A search that is just your name is reported on its own line ("top result for N of M searches on your name"), because ranking first for your own name is not market performance.
On your trend line: Most brands stepped down at this boundary, by 2 to 8 points. That is the truer number; a trend line across the date is not comparable.
- Searches Google answers by location are now spotted from what Google showedNovember 11, 2026
For a local business, some searches get a different Google page in every town. A search like "custom pool builders" shows pool builders near whoever is searching. We cannot read those from outside your town, so we measure them on Google Maps. We used to pick those searches from the words alone. Now we also look at what Google showed: the map results, and whether two separate reads got the same page.
On your trend line: No earlier scan is changed. From a local or regional business's next scan, a few searches may move between the Google list and Maps. National businesses are not affected.
- A search that names one of your towns is checked from that town, every timeSeptember 27, 2026
For a local business we check each search that names one of your towns from that town on Google. Until this date, a town Google does not list under the name you gave us (like "North Houston") was checked for the whole state instead. We now check every town against Google's own list first: a part of a large city is checked from that city (North Houston from Houston), and a nearby town you named in your aim consult (like Spring beside The Woodlands) is checked from that town when a search names it. A local business's searches that name no town, like "chiropractor near me", are now checked from its home town instead of the middle of the state, because Google answers them for where the searcher is.
On your trend line: No earlier score is changed. Only a local business's searches move, from their next scan: those that name such a town, and those that name no town at all. Regional and national businesses check every search from the same place as before.
- A Google answer that came back short is asked for once moreSeptember 25, 2026
For each tracked search we ask our data source for Google's first 100 results. On some days the answer comes back with only the first page or two, even though Google has more. From this date, when an answer holds fewer than 50 results and stops short of what we asked for, we ask once more before scoring the search and keep the fuller answer. A search whose answer stops at the same place both times is one Google itself answers briefly, and is treated as before.
On your trend line: No earlier score is changed. Until this date a short answer left that search out of the part of the score that asks whether you rank at all, so on a day with many short answers the score could sit a little high; after this date those searches are measured, and a brand whose last scan fell on such a day can read a little lower on its next scan with no change in its market. A drop of a point or two at this date is the instrument catching up, not the market.
Losing questions
The order of the questions you are losing carries its own version, because it can change while your score does not. These are the changes to how that order is worked out.
- A losing question now counts Google AI Modev4 · November 3, 2026
How much a losing question is worth depends on how many of the assistants you lost it on, out of those we asked on that scan. Google AI Mode is now one of them when it answers, so on a scan where it answered, the worth of a question reads a little differently.
On your trend line: Your visibility score has its own note for this date. A question you lost on every assistant we asked is worth the same as it was. Other losing questions may read a little differently on a scan where Google AI Mode answered.
- Losing questions are ranked by what they are worth times your chance of winning themv3 · September 9, 2026
Two changes. First, a question counts as one to win whenever the assistants had room for a name nobody asked about and did not use it for you, even if you are already named in some of the answers; before, only questions where you were never named counted. Second, each question now carries two numbers: what winning it is worth (the old impact number, unchanged) and your chance of winning it, which combines how often the answers had room and how often the assistants already name you on questions of the same kind. The list is ordered by worth times chance, so a valuable question the assistants have never named you on sits below a smaller one they already half-know you for. The chance term borrows one question’s worth of answers from your whole service line when a kind has few answers, so one lucky mention cannot swing it.
On your trend line: Your visibility score and its trend line are unchanged. The order of your losing questions, and which questions appear at all, are not comparable to before this date; a question’s worth number is the same scale as the old impact number.
- Losing questions now count whether a slot openedv2 · September 2, 2026
The order of your losing questions now counts whether the answers left room for a third name. When a question names a rival by name, the only way you could have appeared in it is if the assistants brought in a name nobody asked about, so we now measure how often that happened and rank the question by it. A question whose answers never bring in a third name still shows, and now sits last.
On your trend line: Your visibility score and its trend line are unchanged. A losing question’s impact number is not comparable to the one it had before this date.

