September 2026: Google Business Profile, brand monitoring and sector demand

Google approved PivotSEO's access to the Business Profile API. Your business listing is now tracked, archived and analyzed inside the app, and you can reply to your reviews without leaving PivotSEO. The month then went well beyond local: a Watch tab keeps an eye on what is being said about your brand in the press and on the web, a proof of identity finally stops your client being confused with its namesakes, and a project's market can now be set to any of 32 countries. Here is what is new or changed, section by section. The detailed articles are listed at the end.

Connection and settings

• New Google permission « Manage your Google Business Profile listings »: Reconnect Google button in Settings, with a warning as long as the permission is missing.
• Listing selector « Google Business Profile (official API) » in Settings: pick among the listings of the connected account; the choice survives deployments.
• Project city and region editable in Settings, with the Use this city button that copies the listing's city (used for local searches, the heatmap and competitor reviews).
• Dedicated Google account for the listing (September 13): when the listing belongs to a different account than Search Console and GA4, a second button connects that account with the « listing » permission only. Status, switch and disconnect in Settings.
• Statistics denied by Google (September 13): if the Performance API answers 403 (unverified or suspended listing, insufficient rights), the page shows the likely cause and « unavailable » indicators instead of zeros.
• Several listings and several Google accounts per project (September 15): a client with one business per city adds each of its listings in Settings (up to 10, with a search by name, city or address) and finds them in the « Listings tracked by this project » list, each read with the Google account that manages it. Up to 5 Google accounts can be added for listings, on top of the main account. Listing tabs (name and address of each location) appear at the top of the Google listing, Reviews and Heatmap pages; on Google listing and Reviews, the first All listings tab combines the statistics and reviews of every location, with one line per listing and a comparison table. The Sunday scan refreshes every listing. The weekly email has one block per listing (with a direct link to it); in custom reports, a Google listings overview table compares the locations, then each local section is detailed listing by listing, with its own AI commentary.
• Analyses in the project's language (September 15): on Local Presence, AI analyses (period, searches, reviews, competitors) and checklist tips follow the project's language — a project set to English gets English analyses even if your interface is in French.

Google Business Profile page

• Periods of 30, 60, 90 days or custom, and comparison with the previous period or the same period last year: every KPI shows its variation.
• Sticky horizontal menu to jump between sections.
• AI analysis of the period: insights and prioritized actions.
• Daily charts with project events (scans, publications) marked on the curve.
• Listing → site funnel: website clicks then GA4 sessions, engaged sessions and conversions, thanks to the suggested UTM link; warning if the listing points to a URL shortener.
• Seasonality and action mix (website, calls, directions) over 18 months.
• Competitor benchmark: rating, review count, photos and posts versus same-category competitors in your area.
• Breakdown Search / Maps, mobile / desktop and weekday, plus the 12-month trend.
• PDF export of the whole page with a header (listing, period, date).
• Statistics stop at the last day published by Google (about a 6-day delay) so false zeros no longer appear.

Searches that surfaced your listing (new section)

• Brand / competitor / service classification (typos recognized), detected places and « near me » queries, with the share of each kind.
• Cross-referenced with the site (GSC), local volume and captured share, tracked keywords, AI Overviews and the 12-month trend; « new » and « lost » tags.
• Track, Content (suggested page sent to the content queue) and AI (the search becomes a prompt tracked in ChatGPT, Perplexity and Gemini) buttons.
• Listing only block (queries with no page on the site, with a suggested page; meaningless suggestions filtered out) and Site only block (queries where the listing does not appear, with a geographic diagnosis).
• AI analysis of the searches and alerts (brand drop, rising competitor, new city).
• The latest complete month published by Google is used (about a 3-week delay) and shown under the title.

Reviews tab

• All reviews of the listing are fetched and archived: no more 50-review limit, and reviews deleted on Google's side stay in the archive.
• Review health: rating and trend, new reviews per month, reply rate, median reply delay, alerts (negative unanswered for 48 h, rating drop).
• Semantic AI analysis over all reviews, with themes, correction indicators and a correction plan; Retry button on error.
• Priority queue with AI-drafted, editable replies.
• Publish on Google, Edit and Remove from Google: replies are published straight to the listing after confirmation, by the Owner and Editor roles, and synced into the archive.
• What your competitors' customers say: analysis of a competitor's reviews, picked from the list, typed by name or pasted as a Google Maps link.
• Customer vocabulary: the real words from reviews, sendable to the content queue.

Local heatmap

• The grid is centered by default on the connected Google listing (name, address, coordinates); the Other place button analyzes another point.
• If Google does not provide the listing coordinates, PivotSEO geocodes its address in the right city.

Scan, archive, digest and reports

• The listing is included in the Sunday weekly scan and served from cache during the week.
• Unlimited archive in PivotSEO: daily metrics, monthly searches and reviews accumulate at every scan.
• The weekly email digest gets a Google Business Profile block (KPIs week vs previous, new reviews, unanswered negatives, top searches).
• Local category in custom reports: Google Business Profile Performance, Searches That Surfaced the Listing, Google Reviews, Google Listing Checklist and Local SEO (Listings and NAP), with variations; the previously empty Local SEO section is fixed.
• Report footer shows « PivotSEO » by default.

Impact proof, execution and sharing (17 September)

• Impact proof (new page, Home menu): every action you take becomes a dated marker, and PivotSEO compares the same measurements over equal 30, 60 and 90-day windows before and after — AI capture rate, GEO score, Google clicks and position, GA4 sessions. Marking a recommendation as done creates the marker automatically; you can also log work done outside PivotSEO. A window that has not elapsed shows the days remaining rather than a premature result.
• What AI traffic brings in (Google Analytics page): sessions, engagement, conversions, revenue and value per session for traffic from ChatGPT, Perplexity, Gemini or Copilot, compared with organic traffic and the whole site. Revenue only appears if your GA4 property actually measures it.
• Place a schema on the site (Schema Injector, new tab): PivotSEO writes the JSON-LD into the target WordPress page itself, then re-reads the public page to confirm it is there. The Check all button re-verifies placed schemas — a theme, a cache or a plugin can remove one without warning.
• Shareable report link (Reports page): a read-only link in the project's colours, to hand to your client with no account to create. Expiry of your choice (90 days by default), immediate revocation and a view counter. No costs, settings or other projects' data are visible.
• Weekly email: it now goes out on its own every Monday at 9 a.m., owner addresses are resolved correctly, and every email carries an unsubscribe link. The Digest page shows whether the sending service is configured and how many projects will receive the email.

New data sources (17 September)

• 25-week field trend (Core Web Vitals page): the history Google publishes for your origin, per metric and per device — the snapshot told you where you stand, the curve tells you whether the technical work paid off.
• Wikipedia in detail (Entity & Brand page): monthly views of your article over 12 months, page solidity (size, incoming links, last edit) and above all the articles whose text already cites your brand — these are what models read about your sector.
• ChatGPT with web search (Settings → AI engines): until now the ChatGPT engine answered from memory, citing no source. Enable the option so it searches the web before answering, as it does for a real user: PivotSEO then collects the URLs actually cited, which feed Source Authority. Off by default, as each call costs roughly ten times more.

Brand monitoring (18 September)

• New Watch tab, first in External presence (/dashboard/watch): before measuring where the brand is cited, you watch what is being said about it.
• You register the expressions to watch: the company's full legal name, product names, executives and spokespeople, running campaigns. Up to 25 expressions per project.
• Those expressions are collected automatically during the Sunday scan, and on demand with the Collect now button — handy the day before a client meeting.
• Sources covered: the press (Google News, free), Hacker News and Bluesky (free), then the web and Reddit through Google search. Those last two are billed, hence the Include web and Reddit checkbox, unticked by default.
• Instagram, Facebook, X and TikTok are not covered. Their public search is either closed or resold for several hundred dollars a month: better to say so plainly than to let anyone believe the monitoring is complete.
• Every hit you have not opened yet carries a New badge, and the badge is per person: when two agencies work on the same project, the one reading its hits does not clear the other's badge.

Telling the brand apart from its namesakes (18 September)

• The problem: a common name (« Mansour », « Orange », « Olivi ») pulled press articles and web mentions that were about someone else entirely — a public figure, another company, an ordinary word. A name on its own proves nothing.
• The fix: ticking definitely refers to my client on a Watch expression turns it into an *identity anchor* — typically the full legal name, a registered trademark or an executive's name. The whole External presence section then uses it as proof.
• Three levels replace the old yes/no: confirmed (an anchor appears in the text), to verify (the name appears near a context term, but nothing proves it is the client) and rejected. Results to verify are shown separately behind a toggle and count towards no total.
• A Not my client button on every result: the domain is permanently excluded from all of External presence, and cached analyses are purged immediately — the correction shows up right away, without waiting for the next scan.
• As long as no anchor is registered for a brand with a common name, a warning reminds you: on that kind of client, it is the very first setting to get right.

Project market: 32 countries (18 September)

• A project's country could only be picked from six. There are now 32, across four continents, grouped by region under Settings → Project market.
• Picking a country automatically sets the search market and suggests the dominant language; you keep control of both.
• That one setting drives everything else: Google positions, search volumes, domain analysis, the press edition queried, the directories suggested.
• Fix: in Location (search volumes), the country selector of the fine-grained search reset to Canada on every render — so searching for a US state never returned anything. It now follows the project market.
• Press source classification works everywhere now: a German, Spanish, Japanese or Brazilian title is recognized as a news outlet just like a Quebec daily.

External presence: one consistent frame (18 September)

• Press, Web mentions, YouTube, Reddit and Social buzz now all apply the same three settings: the project market, the project's SEO keywords and disambiguation. One page can no longer count what another rejects.
• On YouTube especially, a video that merely says the name with nothing tying it to the client no longer counts as a win.

Web mentions: new indicators (18 September)

• The page used to show only a sentiment score that was usually empty. It now answers the useful question, with four indicators: Sites talking about you, Already picked up by AI (how many of those sites ChatGPT, Perplexity or Gemini already cite for your prompts), High-authority sources, and tone — shown only when the data actually exists.
• Why it matters: landing a mention on a site the models already cite is worth far more than ten mentions on sites they ignore. The Who cites you list puts those sites first: they are your next PR targets.
• When the content analysis API does not cover the brand, the page falls back on Google search results rather than staying empty — and it tells you which of the two sources it is showing.

Flagging a result that is wrong (19 September)

• In Press, every article carries a Flag button offering two choices: « This result is not about my client » (drops that single page) and « This outlet never covers my client » (drops the whole domain).
• Why two scopes: excluding an entire domain because one of its articles was about a namesake threw away the legitimate coverage from that same outlet.
• The exclusion is permanent, applies to all of External presence, and the cache is purged straight away — the correction shows up immediately.

Tracked keywords now drive the search too (19 September)

• The press search fires three extra queries crossing the brand with the project's SEO keywords.
• The gain is a matter of coverage: Google News ranks on the headline, and plenty of articles only name the brand mid-text. Those queries are what surface them.
• Every article now shows the settings that got it kept (identity anchor, tracked keyword, competitor, city), and a framing line states what basis the search was run on.
• The project's competitors count just as much: an article citing your client next to a competing gallery is talking about the right market.

Filtering the watch list and exporting it as PDF (19 September)

• Every watched expression is now a filter: clicking it shows only its hits, several can be combined, each one shows its result count, and « Show all » takes you back.
• PDF button: prints or saves the list exactly as displayed, filters included, with a header recalling the project, the date, the source and the expressions kept.

Saying a source is good, not only bad (19 September)

• On Web mentions, every source carries two actions: Good source and Not my client, plus Undo to reverse either one.
• A confirmed source has its results marked « confirmed » by default, instead of going back through « to verify » every week.
• A deliberate limit: a term you have banned still drops the content, even from a confirmed source. The verdict says the site talks about your client, not that everything it publishes should be accepted.
• A summary line recalls how many sources are confirmed and how many are excluded.

External presence history (19 September)

• New History tab, placed last in External presence (/dashboard/presence-history).
• One card per source (press, entity, Wikipedia, web mentions, Reddit, YouTube, social buzz) with a trend curve, the period covered, and a date picker to reopen any past reading exactly as it was computed that day.
• Those readings were already archived at every Sunday scan, for months; what was missing was a way to read them.
• A gap in a curve stays a gap: « not measured that day » and « zero that day » do not mean the same thing.

Three years of domain history (19 September)

• In Domain overview, changing the period (6 months / 1 year / 2 years / All) had no effect: Google was never asked for more than twelve months, and windows longer than that stayed clickable without changing anything.
• Three years are now requested. A window longer than the published history is greyed out, the number of readings appears next to the selector, and the subtitle says when the curve is already showing everything there is.

Brands and outlets in every writing system (19 September)

• Topic analysis held up in Latin script and degraded elsewhere, silently: an Arabic project extracted no industry terms, a Japanese identity expression written without spaces went unrecognized, and all Asian and Arabic press outside the few well-known titles fell into « Other publications ».
• Fixed for Japanese, Chinese, Korean, Arabic, Hebrew, Russian and Greek, on top of the Latin-script languages. Ten languages checked end to end.
• What it changes for you: a client tracked outside Europe and the Americas gets the same quality of analysis as a Quebec client.

Error message fixed on the domain view (19 September)

• The page blamed the DataForSEO credentials as soon as a domain had no rankings, even when the account was answering perfectly.
• Two situations are now told apart: the API not answering (red message, technical cause) and the domain simply having no rankings in the market queried (amber message, pointing you to the country setting).
• The country almost always explains the gap: the same site can have two keywords in one market and a hundred and fifty in another.

Knowing when a measurement never happened (19 September)

• The automatic Sunday scan could fail without anyone finding out: the trace of the failure lived in memory and vanished at the first redeploy. By Monday morning the incident had never existed.
• Scan attempts are now kept. The dashboard warns you when the last successful measurement is more than ten days old, naming the date — and, when the cause can be identified, the action to take: reconnect Google if the authorisation expired, check an API key or a quota, relaunch a scan that ran out of time. When the cause stays unknown, the page says so rather than sending you off to repair something that works.
• An email goes to the project owner after two consecutive failures, and never more than one a week for the same project.
• Why it matters: flat charts because nothing is being measured look exactly like flat charts because visibility is stalling. It is the costliest failure in the product, and it was the only one you could not see. A project that has never been scanned, on the other hand, is not treated as a failure.

Backlinks: profile, trend and competitive gap (19 September)

• The Backlinks page has been rebuilt. What it shows now: domain authority (out of 1000), the dofollow/nofollow split, twelve months of links won and lost with the monthly net, anchors grouped by family with an over-optimization warning, and a competitive gap — the domains linking to your competitors but not to you, with the list of competitors each source already cites.
• Calculation fix: the dofollow share was invented whenever DataForSEO had not recorded the link attributes. A domain that was never measured displayed as 100% dofollow, which presented beautifully in a meeting and meant nothing. The page now writes « not available ».
• The full detail is in the article Link profile: authority, trend and competitive gap.

Anonymized sector benchmark (19 September)

• What it is: knowing how many of the sources AI engines cite in your sector already cite you. Your Sources pages tell you who cites you; they cannot tell you what is normal for a furniture maker in Quebec.
• Off by default on this instance. That is the main thing to know: the feature only switches on if whoever operates PivotSEO decides so explicitly. While it is off, nothing is aggregated and nothing is computed.
• The safeguards, when it is on: no benchmark is computed below five projects in the same sector and the same market, any project can opt out of the aggregate from its settings, and what is never shared stays never shared — no project name, no domain, no prompt. Only domains cited by the models travel, and those are public by nature.
• Dedicated article: Sector benchmark — what normal looks like in your sector.

No more blank screen without an explanation (19 September)

• Forty-four analysis pages could go completely blank when a call failed. Nothing told a breakage apart from a legitimate absence of data, and the empty screen suggested the tool simply did not work.
• Three distinct messages replace it: a breakage, with its cause shown and a Retry button that replays exactly the same call; a normal absence of data, in a neutral tone, because that is an answer and not a failure; a missing setting, with the exact path to what needs fixing.
• The third case is the most common one: a connection or a setting that is not there, which nobody could guess from a white page.

Your sector's demand, measured by Google (22 September)

• The sector benchmark is no longer silent. It waited for five comparable projects before saying anything — a necessary threshold, but one that condemned it to silence. It now leads with your sector's demand curve over twelve months, taken from Google Trends across a whole category. No other customer's data goes into it: it works from your very first project.
• All sixteen of your sectors are mapped to a Google category. It is always displayed on screen so you can judge whether the comparison applies to you. Four sectors that Google carves up differently carry a « partial match » warning stating exactly what it leaves out.
• The strongest weeks of the year are listed: that is your editorial calendar. Publishing six to eight weeks before a peak gives Google and the answer engines time to index the page.
• Impact proof finally compares against something. Each action now carries, next to its own change, your sector's movement over exactly the same dates, and the gap in percentage points. An 8% rise in a market falling 6% is worth 14 points, not 8 — and an 8% rise in a market rising 12% is really a decline.
• When the benchmark is missing, nothing is shown. A gap computed from one or two weeks would give a coincidence the appearance of a fact.
• Content briefs now carry a date. When the sector has a marked season, the brief shows "publish before …" — six weeks ahead of the peak, the time Google needs to index the page. The date follows into the Markdown export handed to the writer, and the generated article takes it into account: if it is too late to target the peak, the instruction flips and asks for content that lasts, with no wording that would date it. A sector without a season gets no date.
• Detailed articles: Sector benchmark, Impact proof and The Content Workshop.

Your client against their market (23 September)

• The sector benchmark finally places the client, not just their market. Two cards now sit below the demand curve. They use only the project's own Search Console data, set against public Google statistics: nothing comes from another subscriber.
• Your visibility against the market. Your Search Console impressions, week by week, laid over your sector's curve. Three figures: the gap to the market over the year, in points (+8% while the market falls 6% is +14), the same gap over the last eight weeks, and the link to the season, which tells you whether a dip comes from the season or from elsewhere. It answers « is it my fault or the season? ».
• What your market searches for most. The category's 25 most searched queries in Google, matched against your Search Console queries: how many you get impressions on, how many are in the top 10, and how many are to push (visible but outside the top 10 — the quickest wins). Every row leads to a brief.
• Searches for your own brand don't count. « mansour rugs » in 3rd place measures the client's awareness, not their place in the rug market. Those queries are excluded, and the screen says how many.
• The Google category is broader than your offer. Queries where you don't appear are folded below the list so you can pick what concerns you: « clock » at a rug retailer must not inflate the figures.
• Fast-rising queries only show when they touch your vocabulary. Raw, the list mixes trivia with real opportunities.
• The demand curve reads better: a labelled axis (it doesn't start at zero, and a 76-to-100 swing no longer reads as a collapse), months under the axis, each week's value on hover.
• Detailed article: Sector benchmark.

Your impressions in Google's AI answers (26 September)

• New card on the Search Console page: the share of your Google impressions that come from AI Overviews and AI Mode, its four-week trend, the pages Google's AI uses most, and the breakdown by country and device.
• Why it matters: it is the only official measure, from Google itself, of your presence in its generated answers. And an appearance in an AI summary often produces an impression without a click — a page losing clicks may actually be gaining visibility.
• Why it has to be imported by hand: Google has published this report in Search Console since June 2026, but hasn't opened it to its API. No tool can read it automatically. One export a month is enough; PivotSEO will fetch it on its own as soon as Google opens it.
• Where the figure shows up next: a tile on the home page, an indicator in the summary of client reports and the weekly email, and an « Impressions in Google's AI answers » section in custom reports. An import older than 35 days no longer shows there: an old figure would pass for this week's.
• Import history: every uploaded export is kept in full, with its original CSV files (downloadable again as a .zip). Two imports can be compared: AI displays per day, the site's AI share, and the pages moving the most. Exports of different lengths are compared as daily averages, never totals.
• Detailed article: Search Console's Generative AI report.

Compare every section with the past (26 September)

• A period bar at the top of sections: pick a period (7 days, 28 days, 3 months, 12 months or dates) and compare it with the previous period, the same period last year or dates of your choice. Your choice follows you from one menu to the next.
• Good to know: you can only compare with the past already recorded. History starts when your project is created and grows as long as it stays active — after a year, you compare with last year. Search Console (16 months), keyword volume (12 months) and the Google Business Profile already have a past; everything else builds up over time.
• Detailed article: History and comparisons: how it works.

Google's AI report crossed with AI Overview and Sources (26 September)

• AI Overview → « Google's official measure »: side by side, PivotSEO's measure (on how many of your tracked keywords you're cited in the AI summary) and Google's (the share of your Google displays that happen in an AI answer). Two different measures: a sample of keywords checked from one location, versus every real search, without detail.
• Your cited pages: for each of your pages cited in an AI summary, the keywords it's cited on, and its real AI displays according to Google.
• Keywords to add to tracking: pages shown a lot by Google's AI but never seen in your analyses — they rank on searches you don't track.
• Sources → « Your pages seen by AI »: your pages in three groups — used by Google AND other AIs (to protect), only by Google, only by other AIs (ChatGPT, Perplexity, Gemini, Claude).
• The crossing is page by page: Google's report doesn't give queries. The report must be imported on the Search Console page.

Your data keeps its history (26 September)

• Search Console, query by query and page by page: a snapshot is kept at every weekly scan (and at your first visit of a day). On the Search Console page, the Compare two dates card shows the queries or pages that moved most, with new and lost ones. Snapshots of different lengths are compared as daily averages. GA4 and Bing are recorded the same way.
• Keywords: volume, difficulty and cost per click of your tracked keywords are kept date by date. The keyword page shows their evolution.
• Recommendations: every status change is dated, with who made it.
• AI Overview: on the Google tab, Compare two dates shows the keywords where your brand gained or lost a citation, and the sources gained or lost.
• Nothing gets lost anymore: large snapshots were skipped by the archive; they are now compressed and kept. And the External presence history kept the oldest snapshots instead of the newest beyond 400: fixed.
• These histories start today: at least two snapshots are needed to compare.

Your clicks vs Google's AI answers, query by query (26 September)

• The card explains itself first: on many searches, Google shows at the top a summary written by its AI, with links to the sites it used as sources. The sites shown next to a query are those sources — cited instead of you.
• Four figures: the citation rate (the share of your clicks, on queries with an AI summary, coming from queries where Google cites you — the number to raise), clicks to defend, estimated lost clicks (gap between the clicks a page at your position usually gets and yours — an estimate, not a measurement) and the analysis coverage.
• Each query has a state: cited, not cited, sources unknown, no AI summary, not analyzed yet. Click to see your ranking page, the exact pages cited instead of you (tracked competitors flagged) and a link to create a brief.
• The sites Google cites most instead of you are grouped: your real competitors in Google's answers, not necessarily the ones you track.
• Analyze more queries: 10, 25 or 50 at once.
• Two fixes: a site cited through two of its pages showed twice, and a query whose AI summary had no readable source was counted « to defend » although we can't tell whether you're cited.

Your analyses no longer disappear (26 September)

• An analysis you ran stays there: closing the page, refreshing it or coming back the next day shows the latest analysis, with its date (« Analysis of … »). Nothing is re-run when the page opens: a new analysis only starts when you click the button.
• Every analysis is kept in a history: the list of past analyses lets you reopen one, and the period bar compares the latest analysis of one period with another period's.
• Tools covered: Brand Perception Detector (positioning, tone, adjectives and topics gained or lost per AI engine), page audit, SEO audit, structured data checker, SXO, SERP clustering, GEO content audit, AI answer accuracy, content analysis, content evaluation and meta tag generator. For a URL audit, the comparison shows the score before → after and the checks fixed or newly failing.
• AI answer accuracy no longer re-runs on its own: the page re-queried the engines on opening whenever the result was over seven days old. It now shows the latest check.

Fewer paid calls, data always there (26 September)

• Opening a page no longer triggers a paid call (DataForSEO, AI models) when the data was already collected, even long ago: the page shows it with its date, « Data from … », in amber when it's old. The Refresh button re-collects it when you decide.
• Pages concerned include: backlinks, SERP analysis, domain overview, gap analysis, competitor discovery, search intents, Google reviews without a connected listing, local competitors, Google Trends.
• The Sunday scan refreshes what you look at: data opened at least once is updated every week (or every month for monthly data). A page nobody opens costs nothing.
• Even the very first opening of a page with no data no longer triggers anything: a Run analysis button lets you decide (see « Figures that match » below).

Search Console reorganised (26 September)

• A summary at the top: clicks, impressions, CTR and position compared with the chosen period, then the share of your displays in Google's AI, your citation rate, your clicks to defend and the traffic value. Each figure leads to its section.
• Two distinct sections: « Classic Google search » (charts, comparisons, opportunities), then « Google and AI » (Generative AI report, clicks vs AI answers). Two links at the top jump straight to either one.
• The « All queries » list is at the bottom: it's a digging tool, not a summary.
• Totals match: the cards and the chart cover the same days and the same site totals. Per-query tables state that they exclude queries Google anonymises, which is why they add up to fewer clicks.

A much faster dashboard (26 September)

• The dashboard fired about twenty requests at once, which queued in front of the database: about 8 seconds before the figures showed. It now sends a single one for the top of the page, then the rest in small waves: figures show in 1 to 2 seconds.
• Right after a PivotSEO update, the first display may take a few more seconds while data reloads.

Your analyses keep running in the background (26 September)

• Start an analysis, then go elsewhere: brand perception, audits (page, SEO, GEO, structured data), AI answer accuracy, SXO, SERP clustering, the « clicks vs AI answers » check, brief writing and regeneration, SEO drift check and the local heatmap keep running on the server if you switch tabs in PivotSEO.
• An indicator in the top bar shows « N analyses running » (« 1 analysis running »), with each one's progress, and tells you when an analysis is done, with a link to the result.
• When you come back to the page, progress resumes where it is, then the result shows.
• PivotSEO updates no longer wipe anything: every finished step is saved. If the server restarts during an analysis, it resumes on its own at the next step, without redoing — or re-paying for — what was already done. The heatmap resumes with one click on « Resume ».

Local heatmap: faster, stored (26 September)

• Grid points are queried six at a time instead of one by one, for the same cost: a 5×5 grid goes from about 2–3 minutes to about 30 seconds.
• Every grid is saved and shows immediately when the page opens; past grids reopen and compare from one period to another.
• A point in error no longer wipes the others: it shows « ! ».
• Detailed article: Local heatmap.

More analyses kept (26 September)

• Web Vitals and their advice, per-engine AI advice, traffic analysis, domain overview, Google listing audit, simulator, prompt optimizer, local pages, ASO listing audit and naming advice, content template: the latest result shows when the page opens, with its date and history.
• Briefs keep their versions: every regeneration of a section or the article keeps the previous version, which you can restore.

Figures that match, and no paid call when a page opens (26 September, continued)

• Opening a page never costs anything any more, not even the first time. A section with no stored data shows a Run analysis button instead of starting a paid collection on its own: SERP analysis, backlinks, domain overview, competitor discovery, AI Overview, Google Trends, local reviews, keyword overview, keyword gap, sector benchmark, nearby competitors…
• Data already collected is shown whatever its age, with its date and a Refresh button: you decide when to pay for a new collection. An empty answer (no gap found, a domain with no ranking keyword, a listing not found) is recorded too, so it isn't asked again at every visit.
• Search Console traffic value: it relies on each query's cost per click, which is paid data. It is now computed with one click on Compute the value, no longer at every page load.
• The same figures from one page to the next: the home page, the period tiles and the detailed pages read the same period and the same source. The home page's Search Console clicks are the tile's; the « GEO score » ring and the GEO tile give the same figure. When two blocks cover different windows (28 days and 90 days, for instance), their dates are written.
• GA4: the « conversion rate » could show 983%, because GA4 counts several key events per session. It is now called Conversions per session (9.83). The « Organic clicks » column copied the sessions: it now shows each page's real Search Console clicks, with its top queries. The table pages through 50 rows at a time and can be filtered with a search.
• Alerts: a single « click drop » alert per query over 7 days (existing duplicates removed), with the dates and the comparison baseline in the message. Alert types are written in plain words.
• Google Trends: curves were losing almost all their points, because every week Google flags as incomplete was dropped instead of only the last one. Fixed; a week without a value leaves a gap, never a 0.
• Local competitors: limited to your Google listing's category, period comparisons included. Backlinks: your competitors are recognised. When the compared period has no snapshot, the page writes « no data » instead of « not available ».
• What shows again: stored heatmap grids, audits and Core Web Vitals display again; accents in press titles are repaired; your own site no longer appears among watch findings.
• Project costs: DataForSEO collections started from pages were counted outside the project. They now appear, line by line, in each project's costs.
• Your language's formats everywhere: numbers, percentages, amounts and dates follow the interface language (17.5%, $0.10, 1,234 in English; 17,5 %, 0,10 $ in French).
• Reliability: a double click no longer starts the same analysis twice (nor two billed heatmap grids); an analysis interrupted by an update resumes on its own, three times at most; Search Console, GA4 and Bing comparisons no longer read a day missing from the archive as a traffic drop.

Finishing touches, 26 September

• Home page: the GEO score chart line is back. The Search Console, Bing and GA4 blocks load first and show « Loading » while they read, or the real error if reading fails, instead of a misleading « No data ». The priorities block reuses the last calculation made for the same scan, so it shows at once.
• One set of figures: the AI engines page separates queries sent from usable answers; a query that returned an engine error is not counted in the capture rate. « Missed prompts » are the ones that didn't mention the brand in any answer over the period. Queries at risk from AI are counted as distinct queries everywhere, as on the Search Console page. Comparisons give the exact ratio (« 1.6 times more often, 32 mentions vs 20 »). The number of reviews is the same on every page. Watch counts use the same list as the « All » tab, and its filter is now called « News ».
• GA4: the engagement rate is recomputed for both periods and compared in points (+2.1 pts), rather than as a percentage of a percentage. Countries show in the interface language.
• Alerts: repeated alerts on the same query are grouped in a single row, which lists each occurrence with its dates.
• Project costs: periods stop at yesterday, the last full day; the current day shows separately, on a « Today (in progress) » line.
• Bing: a query also present in Search Console is no longer listed as Bing-only.
• Local heatmap: the latest grid always shows, with its history, whatever period is chosen.
• Local competitors: only businesses whose main Google category matches yours are kept. The others (a furniture chain for a rug store, for instance) go in a collapsed « excluded » list.
• SERP analysis: the page now reads the results page features (AI Overview, local pack, « People also ask », etc.) and your organic position, comparable to Search Console's. An older reading says so and offers Refresh.
• Backlinks: an anchor made only of symbols (« ] », for instance) is classified « Empty / symbol » and no longer triggers a false over-optimization warning. Your competitors are recognised by their domain; for a competitor without one, the page asks you to add it in Settings. The period comparison is faster and offers Retry if it takes too long.
• Google Trends: an incomplete curve is labelled as such, with a Refresh button.
• Search Console: the full query list pages through 50 rows at a time, with a search and column sorting.
• Domain overview: the monthly history uses the same country market as the ranked keywords, and two captions explain why the two counts differ: every keyword of the month on one side, the first 200 of the list on the other.
• Press: opening the tab never runs a search. The page shows the last stored result, with its date; it is collected again at the Sunday scan, from free sources, or when you click the button.
• Formats: a true minus sign (−6.9%), short dates on chart axes (« Oct 5 »), country names in the interface language, abbreviated large numbers (6.6k).
• Faster period comparisons: heatmap, technical audit, domain overview and AI citations no longer reread the whole archive at every display.

Latest fixes, 26 September

• A faster home page: priorities and the summary are prepared right after each scan. When you open the page, the last result shows at once, then it is refreshed in the background.
• Home page blocks give the same figures as their pages: « In Google's AI answers » uses the Search Console report over the selected period and shows the change in points; the Bing block shows clicks, impressions and queries over the roughly six-month total Bing provides; the GA4 block reads the selected period's dates and shows them; « X queries analysed out of N » says where the count comes from.
• Recommendation texts: numbers in your language's format, clicks and sessions over the last 28 days with their dates, share of voice identical to the Mentions tab.
• GA4: channel names are translated. The page shows the previous reading at once, dated, then refreshes it in the background; it no longer slows down the rest of the application.
• Reviews: the number of reviews and the rating are the same everywhere, read on your current Google listing.
• Local competitors, « Local market »: your declared competitors carry a « declared competitor » badge. They come from the project's Settings and are therefore not filtered by Google category, unlike nearby businesses.
• Period comparisons: when the compared period has no reading, the exact dates are written (« no data between 1 and 28 August »). Today's reading describes the state at the close of the period.
• Countries and domains: country names show in the interface language, domains in lowercase.
• Search Console, « Your clicks vs AI answers »: each tab's count matches the rows it lists; « All » counts the analysed queries.
• Backlinks: generic anchors (« View website », « click here »…) are recognised. An anchor counts as your brand only if it contains the brand name as a whole word.
• Local heatmap: if no grid can be displayed, the page says how many grids are stored for the project, rather than just an empty screen.
• Faster pages: pages that load several blocks at once no longer wait for each block to open its database access: these accesses now stay open.

AI analysis of your statistics and removing analyses (26 September, evening)

• New « AI analysis » tab on the Search Console, GA4 and Bing pages, next to the Data tab, with the same period bar. The chosen tab stays in the page address: a shared link or a reload reopens the right tab.
• Claude reads all the page's figures for the chosen period and cross-checks them, citing the figures. For example: a page gaining impressions but losing clicks (its title needs reviewing), or an AI share rising while clicks fall. You get a summary of up to 5 sentences, then three blocks of 4 to 6 points: Highlights, Concerns and Actions. On Search Console, a fourth block, Google and AI, draws on the imported AI report, the queries where you are cited or not, and the AI Overview.
• Generated on demand: the analysis runs when you click Generate the AI analysis (about $0.02). It relies only on figures already stored: no call to Google, Bing or DataForSEO. It is kept and shown again for the same period.
• Regenerate creates a new analysis and never overwrites the previous one. If the period's figures have changed since, a notice invites you to regenerate it. On Bing, queries remain a roughly six-month total, and the analysis says so.
• History and comparison: under the tab, the analysis history lets you reopen or remove an analysis. Compare two analyses puts the key figures side by side with their change, shows both full analyses, and marks each action as new, repeated or dropped. Comparing costs nothing.
• Local presence also has its AI analysis tab: Google listing, reviews, directories, nearby competitors, local pack and heatmap, with a Local competition block.
• Heatmap: remove a keyword from « Tracked keywords »: click the ×, then confirm. Its grids stay stored; running a grid on that keyword again brings it back.
• « Analysis history », everywhere: remove an analysis (× then confirm). It stays stored, but is no longer shown or compared. If you remove the one being displayed, the most recent remaining one takes its place.
• « AI impressions per week » (Search Console): click a bar to see the week's figures — AI impressions, AI share and change vs the previous week. Before, the detail only appeared on mouse hover, impossible with a finger or on a tablet.
• Local competitors: the page showed no positions at all because of a reading error; this is fixed. New block Local searches that lead to your listing (from your Google Business Profile). The local pack of your tracked keywords shows your rank, after the next SERP analysis refresh. The website of a declared competitor with no site in Settings is recognised when possible (« recognized domain »); otherwise, the page invites you: « Add this competitor's website in Settings ».
• A faster application: the server was moved next to the database (in the United States) and database connections stay open. Pages that load several blocks at once are much faster.

AI analysis on every page, in reports, and the Viewer role (26 September, night)

• The « AI analysis » tab reaches every statistics page: Dashboard, Mentions, Sources, Responses, External presence (sector benchmark included), Keywords, Opportunities (cannibalization included), Content audits, Technical audit (Web Vitals included), Competition, Recommendations, Trends, Impact, Timeline, Matrix, AI Overview, AI engines, Rankings, the four App Store pages, Local heatmap, Local competitors and Domain overview, on top of Search Console, GA4, Bing and Local presence.
• Each analysis reads the figures its page shows; a page with tabs has a single analysis covering all its tabs. One more block is added depending on the theme: AI visibility, Competition, Local competition or Google and AI. If the page has no stored figures for the period, generation is refused with a message, at no charge. Pages without a period bar (backlinks, keyword overview) have none.
• AI analyses now go into reports: new AI analyses of the period section in the custom report (choice of pages; the analysis of the exact period, otherwise the closest one with its dates, otherwise « No analysis generated for this period »), a Generate the missing analyses (N) button with the estimated cost, a section in the automatic post-scan report (last 30 days), a What the AI analysis says paragraph in the executive summary and the weekly email, PDF and print. Building a report never generates an analysis by itself.
• Shareable report link: new Include the AI analyses of the period option. Links created before don't include them.
• Viewer role: a viewer can no longer trigger a paid action (Refresh, Run analysis, Generate the AI analysis, Compute the value, Analyze on the heatmap…). The buttons are disabled, « Reserved for editors », and the server refuses the request with a clear message. A viewer cannot remove an analysis or a keyword either.
• Up-to-date comparisons: they refresh as soon as the data changes (no more delay of up to 12 hours) and when you save the project settings.
• Keyword history: Google's published monthly volume and the snapshot taken on the 1st of the month are both kept. Search history: months never rebuilt (May 2026, for example) are rebuilt with Fill in missing months, from Google's free APIs.
• Detailed articles: AI analysis of the period, Team: invite users and set what they can do, History and comparisons: how it works.

Fixes

• Search Console: the CTR column of the query table showed a value divided by 100 (0.27% instead of 26.6%), reused in the opportunities. The page's totals now all cover the same dates, and each table says what it counts.
• Bing: impossible positions (0.4, −0.1) came from a division by 10; an unmeasured position shows « — ».
• Dashboard: two different « GEO score » figures coexisted (average of all weeks vs latest reading); there is only one now, the latest reading.
• Matrix: the GEO score was the same for every keyword; it is now computed keyword by keyword, from the prompts that contain it (at the next scan). The « Diff. » column is now called « Difficulty ».
• Backlinks and cannibalization: the comparison showed « not available » next to visible figures; it now reads the same snapshot as the page.
• Sources: a banner said your site wasn't cited when it was; your site is recognised the same way everywhere (brand domains and Search Console property).
• Score by engine: recommendations are computed on the same period as the cards.
• One period selector and one project selector per screen (Rankings, Local SEO, Impact proof, Watch, Alerts), placed under the title.
• Competitor discovery: YouTube, Facebook, Instagram, Pinterest are set apart (« Platforms »); local competitors: the search uses your Google listing's category.
• Project costs: a daily cost rounded to $0 made the next period look « new ».
• Texts: accents restored, English labels translated (monthly summary, rankings, mentions, alerts), broken press titles repaired, « 100+ » when the press feed is full.
• The Search Console, GA4 and Bing archive recorded nothing. Shipped in July to keep your history beyond Google's sixteen months, it never actually saved a row. Fixed: open each project's Search Console page once and it retrieves sixteen months of history in one go. Without that archive, the « Your visibility against the market » card stays pending.
• « Claude » and « Gemini » answers fetched through DataForSEO were failing: their default models (Claude 3.5 Haiku, retired by Anthropic in February, and Gemini 2.0 Flash) were no longer offered. They are replaced by Claude Sonnet 5 and Gemini 2.5 Flash.
• The scans' Claude engine moves to Claude Sonnet 5 (instead of Haiku 4.5): a model closer to the one Claude users see, so a more faithful visibility measure. Briefs, content audits and ASO analyses also move to Sonnet 5. Expect a slight shift in the Claude score at the next scan: it's the model change, not your visibility.
• The content queue didn't keep what you added, nor status changes or deletions. Fixed at the same time, for the same reason.
• The top of the demand curve showed as a large flattened black blot, cut off by the top of the frame.
• The sector benchmark had two project selectors. The one in the top bar is enough.
• Google Trends returned worldwide trends to every project: a plumber in Laval was reading the planet's seasonality. Curves now follow the project's market.
• The last week of every Google Trends curve was wrong. Google always returns the current, necessarily incomplete week with its value renormalised to 100, so every curve ended on an imaginary surge. That week is now discarded.
• The listing chosen in Settings disappeared after a deployment.
• « This page couldn't load » message on the Google Business Profile page during a deployment.
• Average rating displayed with stray decimals.
• Last days of the charts at zero (Google publishing delay).
• Review reply delay computed as a median instead of a mean, so a few old reviews no longer skew it.
• Truncated semantic analysis on listings with many reviews.
• False positives in brand / competitor detection within searches.
• The dashboard shows the real date of the last scan.