Every week we send the same 20 buyer questions to ChatGPT, Perplexity, Gemini, Grok, and Claude, and record which brands each engine names and which sources it links to. This is the update for the cycle ending August 10, 2026. It is the 19th round since March.
Two items this cycle. The first is a measure we had not previously calculated: across the full dataset, only 3 of the 20 questions have produced the same unanimous answer from all five engines in each of the last five rounds, and 7 have never produced agreement once. The second is a revision. Our previous read on Perplexity's Reddit citation rate does not hold against this cycle's data, and the corrected series and the threshold we now apply before calling a trend are both below.
The Bottom Line
Most of what AI search engines say about a category is unsettled, and the settled parts are settled for a reason you can see in the question itself. Three questions have a locked answer, and all three name a brand inside the question. Seven questions have never once produced agreement across five engines in 19 rounds, and six of those seven name no brand at all. If you are checking your AI visibility on a single engine, on a single day, you are sampling from a set of answers that mostly re-rolls.
Only 3 of 20 Questions Are Actually Settled
We track how often all five engines name the same brand first. The weekly rate for that has bounced between 15% and 45% over the past five rounds with no direction, which makes any single week's reading useless for planning. So we counted something more durable instead: for each individual question, how many of the 19 rounds produced a unanimous five-engine answer.
| Question | Unanimous rounds (of 19) | Last 5 rounds |
|---|---|---|
| best alternative to Mailchimp | 15 | 5/5 |
| best alternative to Google Analytics | 14 | 5/5 |
| Vercel vs Netlify comparison | 13 | 5/5 |
| best CRM for startups | 9 | 3/5 |
| best hosting for Next.js apps | 9 | 1/5 |
| product analytics platform comparison | 6 | 2/5 |
| best email marketing platform for startups | 5 | 3/5 |
| best platform for deploying web apps | 5 | 1/5 |
| best project management tool for engineering teams | 3 | 1/5 |
| best alternative to Jira | 3 | 2/5 |
| CRM software comparison 2026 | 1 | 0/5 |
| what project management software should I use | 1 | 0/5 |
| analytics tools for startups 2026 | 1 | 0/5 |
| what CRM should I use for B2B sales | 0 | 0/5 |
| best alternative to Salesforce for small teams | 0 | 0/5 |
| lightweight project management for startups | 0 | 0/5 |
| what email tool should I use for newsletters | 0 | 0/5 |
| email marketing software comparison | 0 | 0/5 |
| best analytics tool for SaaS | 0 | 0/5 |
| cheapest cloud hosting for startups | 0 | 0/5 |
The three questions at the top all name a brand in the question, and the answer is the brand that was named. Ask five engines for the best alternative to Mailchimp and all five will tell you to look at Mailchimp. Six of the seven questions at the bottom name no brand at all.
The pattern is directional and not absolute, and the table shows both exceptions. "Best alternative to Salesforce for small teams" names a brand and has never been unanimous, though it is not really in contention either: it returned Salesforce four out of five in four of the last five rounds, one defecting engine short of the settled group. Running the other way, "best CRM for startups" names nobody and has still been unanimous in 9 of 19 rounds.
The practical read: the questions buyers ask before they know which brands exist are the ones with no settled answer, and they are also the questions where checking one engine will mislead you most. This week, "best analytics tool for SaaS" returned PostHog, Amplitude, Mixpanel, Amplitude, and Mixpanel. "Lightweight project management for startups" returned four different brands from five engines.
Corrections and Revisions
Revised: Perplexity's Reddit citation rate
Our previous update reported Perplexity's Reddit citations as a sustained decline. That read is revised. The engine returned to 17 of 20 answers this cycle, with 34 Reddit URLs, the highest figure it has recorded.
| Metric | 4 rounds ago | 3 rounds ago | 2 rounds ago | This round |
|---|---|---|---|---|
| Perplexity answers citing Reddit | 18/20 | 11/20 | 9/20 | 17/20 |
| Perplexity Reddit URLs | 18 | 16 | 17 | 34 |
| Perplexity total sources retrieved | 179 | 234 | 244 | 401 |
| Reddit URLs across all 5 engines | 74 | 58 | 74 | 89 |
The full series now reads 17, 18, 17, 19, 18, 11, 9, 17. The two low readings were a dip inside a stable band, not the start of a decline.
This is the second revision of its kind in the program, after a hosting brand's citation lead that held for three rounds before flipping. Two instances of the same measurement error justify a threshold, so we calculated one. Across every brand and engine citation series in the dataset, there are 25 cases where a metric fell for two consecutive readings and a further reading exists to test it. It recovered in 16, continued falling in 5, and held flat in 4. A two-reading decline reverses 64% of the time.
The threshold we now apply, and the one we would recommend for any internal AI visibility report: three readings constitute a streak. It qualifies as a trend only when a subsequent out-of-sample reading confirms it. On this dataset, a two-reading decline carries no predictive value.
The coverage point is separate and holds regardless. The all-engine total rose while one engine's figure fell. Measured on Perplexity alone, this window would have read as a channel collapsing. Reddit citations across all five engines reached a record 89 this cycle.
Corrected: citation matching methodology
This cycle's collection initially recorded Google Analytics earning its first citation in 19 rounds, which would have ended the longest zero-citation streak in the dataset.
The result was an artifact. Our domain matcher tested whether a brand's domain and a source URL's domain contained one another in either direction, so any google.com URL registered as a citation for analytics.google.com. Two retrieval artifacts of the form google.com/search?q=time+in+Atlanta triggered it.
The rule now requires an exact domain or subdomain match. We re-scanned all 19 rounds, corrected the affected record, and confirmed no other round was affected. Google Analytics remains at zero citations across 19 rounds while being named by all five engines nearly every week, the pattern covered in mentions without citations.
Both items appear here because the corrections are part of the dataset. A figure is only as good as the method behind it, and readers who use these numbers should be able to see when a method changed.
Also Worth Noting
- Perplexity retrieved 401 sources this round, up 64% in two weeks and up from a 143-to-181 range that held for its first 16 rounds. Its citation count did not move with it, holding at 15. We are not drawing a conclusion from this yet. By the rule above, it needs another reading.
- ChatGPT produced 13 brand citations, its lowest in 19 rounds, from a retrieval pool that did not change size. That is a selection change, not an availability one.
- Four brands were named by all five engines and cited by none: Google Analytics, Render, Monday.com, and Railway, 48 mentions between them and zero links. They span every size tier, which rules out company size as the explanation.
- Project management produced 6 citations from 62 mentions, the weakest ratio of any category for the fifth round running, while also producing two of the six unanimous answers this week.
- Two brands were invisible to all five engines for the second round running. One has been named exactly once in 19 rounds.
- Total citations came to 71, inside a nine-citation band that has held for five rounds. Underneath all the movement above, the overall volume is the most stable number we track.
What We Are Watching Next
Whether Perplexity's retrieval pool stays above 350 sources. If it does, that is a platform change worth writing about. If it falls back under 250, it was a spike, and we will say so here.
Methodology
20 natural-language buyer questions across 5 B2B SaaS categories (CRM, project management, email marketing, analytics, developer tools), covering 25 brands tagged enterprise, midmarket, or startup. Each question goes to ChatGPT, Perplexity, Gemini, Grok, and Claude through live API calls, not cached results, on a weekly cycle since March 2026. This is round 19.
We record which brands each engine names, the order it names them in, and every source URL it returns. A brand counts as "cited" only when a source URL on that brand's own domain appears in the response, and as "mentioned" when it is named without one.
This round returned 99 of 100 engine-question runs. Gemini failed one question after repeated rate limits. A question counts toward a unanimous total only when all five engines returned an answer naming a tracked brand, which is why "cheapest cloud hosting for startups" rarely qualifies: it routinely produces answers naming no brand we track.
Two limitations worth stating. We wrote these 20 questions ourselves, so the question-shape pattern above is a variable we selected and then measured, not one we discovered independently. And we only observe outputs. We have no visibility into what any engine does internally.