Siftly Review 2026: The One That Publishes Its Sampling

We have written, in five different articles, that only Evertune publishes how many times it runs each prompt. That is no longer true, and it was already not true when we wrote it most recently. Siftly publishes a response count per plan alongside a prompt count and an engine list, which is enough to derive the runs per prompt with two divisions. It also states that it reports mention rate with confidence intervals.

Those are the two things this site has spent months arguing the category owes its buyers. One competitor does both.

Disclosure: EchoWi competes with Siftly. Every price, prompt count, response count, engine list and quoted claim below was read from Siftly’s own pricing page and its ChatGPT visibility page on 6 August 2026, and the arithmetic is shown so you can repeat it.

The short version

1. From $79 a month, with a genuinely free tier and a seven-day trial on every plan.
2. It publishes responses tracked per month, which almost nobody does.
3. Divide and you get the runs per prompt: about one a day on Starter, about three on the $599 tier.
4. It reports confidence intervals, in its own words, “so a drop is a signal rather than a coin flip”.
5. It bundles content too, from 8 to 60 blog posts a month, which is the shape of half this market now.

What it costs, verified

Annual billing takes 20% off, paid upfront. Seats run 1, 1, 3, 5 and unlimited. Historical data runs 30 days, 90 days, 6 months and 12 months, which is itself a useful signal about what each tier is for.

The engines build up: ChatGPT and Google AI Overviews on Free, plus Perplexity on Starter, plus Gemini on Growth, plus Copilot, Grok and DeepSeek on Scale. Claude is Enterprise only, which is the same line almost everyone in this market draws.

The division nobody else lets you do

Here is why the response column matters more than anything else on that page.

A response is one answer from one engine to one prompt. Every tool consumes them; almost none publishes how many it buys you. Siftly does, and the prompt count and engine count are right beside it, so:

Two divisions. That is the entire method, and it answers the question we have asked every vendor in this comparison and been refused by all but one.

It also sets the honest expectation. Starter is a single daily run, which by our own arithmetic is a sample of one and cannot support a rate. The $249 and $599 tiers buy two and a half to three runs a day, which over a month is 75 to 90 observations per prompt-engine slot, and that is where a per-prompt number starts to mean something.

What those run counts actually buy

Our piece on share of voice works through the arithmetic: a rate near 30% measured over 100 runs carries a 95% confidence interval of about plus or minus 9 points.

Scale gives roughly 90 runs per slot a month, which lands almost exactly there. So a Siftly Scale customer looking at a 30% mention rate for one prompt is looking at something between 21% and 39%, and aggregating across 150 prompts brings the set-level figure into genuinely usable territory.

Starter’s one run a day gives about 30 a month per slot, which is plus or minus 16 points. Better than a single check by a wide margin, and not a number to make a decision on prompt by prompt.

None of that is a criticism. It is the first time in this comparison that we have been able to state what a tier’s numbers are worth, and we could only do it because the vendor published the inputs.

The confidence-interval claim

Siftly’s own page states: “Siftly reports a mention rate per model with confidence intervals, so a drop is a signal rather than a coin flip.” It also says “each prompt runs many times daily” and heads a section “Why one ChatGPT check isn’t enough.”

That is the argument this site has been making, from a vendor, on its own marketing page. We have no way to audit the intervals themselves without an account, and the page does not state which interval method it uses or what confidence level, so treat this as a published claim rather than a verified one. It is still a claim almost nobody else makes.

What an answer costs here

$599 ÷ 108,000 = $0.0055. That is the cheapest answer in this comparison by some distance, and it is cheap for a structural reason: Siftly is selling repetition rather than breadth. A hundred and fifty prompts is not a large panel. Running them three times a day on eight engines is a lot of collection.

Which is the right trade for measurement and the wrong one if you need coverage of a broad category. Decide which you are before comparing the price to anything.

It also writes, which is now the norm

Starter includes 8 blog posts a month with auto-publish, Growth 30 with FAQ and schema markup, Scale 60 with multi-language. There are outreach campaigns too: 5, 10 and 30 a month, plus Reddit engagement from Growth.

That puts Siftly in the same category as Cognizo and Writesonic: measurement bundled with production. Sixty auto-published posts a month at $599 is a volume play, and the same caution applies as everywhere else in this market. Auto-publishing content aimed at a prompt set is a bet that volume beats specificity, and nobody has published a controlled test either way.

The difference here is that the measurement half is the strong half, which is not true of every hybrid.

Where Siftly is the right call

You want a number you can defend. On the $249 or $599 tiers you get run counts you can derive, intervals the vendor says it reports, and a defined panel of 100 to 150 prompts. If your category is narrow enough for that panel, this is the most statistically serious offer with a published price in our comparison.

The free tier is also a real one: 10 prompts on 2 engines, refreshed weekly, 100 responses a month, no card. That is enough to find out whether your category produces AI answers at all, which our own measurements suggest is the first question to ask.

Where it is the wrong call

You need Claude, which is Enterprise only and unpriced. You need a broad panel, because 150 prompts is the ceiling below Enterprise. Or you are on Starter and about to treat one run a day as a rate, which is the mistake this whole comparison exists to prevent.

What we could not verify

• The engine count on Growth. The plan names four and the page’s own engine grid appears to show five. If it is five, Growth’s run rate is 2.0 a day rather than 2.5.
• The interval method and confidence level behind the confidence-interval claim.
• Whether Siftly queries web interfaces or APIs. Not stated on either page we read.
• The Enterprise price, and its refresh frequency.
• Whether responses are distributed evenly across slots or weighted toward some prompts. Our runs-per-slot figures assume an even spread.

The correction

Five of our articles say some version of “only Evertune publishes its sampling”. Superlines, Quattr, Semrush, Writesonic and Surfer all carry it. Those sentences are now corrected to name both.

It is worth saying how the error happened, because it is the failure mode this site warns other people about. We built the catalogue by reading pricing pages of vendors we already knew, and Siftly was not in that set. It arrived through a different door: we asked four AI engines which tools measure share of voice, and Siftly was among the domains they cited. A comparison assembled from a list you already have will always be missing whatever is not on the list.

Every price here sits alongside the rest, each read from the vendor’s own page, in our comparison of every tool we verified.

Common Questions About Siftly

How much does Siftly cost?

$79 a month for Starter, $249 for Growth and $599 for Scale, with a free tier at $0 and Enterprise at custom pricing. Annual billing takes 20% off, paid upfront, and every plan carries a seven-day trial with no card on the free one.

How many times does Siftly run each prompt?

It does not state the figure directly, but it publishes the inputs to derive it. Dividing responses by prompt-engine slots gives about one run per slot per day on Starter, 2.5 on Growth and 3 on Scale. That is more than any other vendor in this comparison discloses except Evertune.

Which AI engines does Siftly track?

ChatGPT and Google AI Overviews on the free tier, plus Perplexity on Starter, plus Gemini on Growth, plus Copilot, Grok and DeepSeek on Scale. Claude requires Enterprise.

Does Siftly report confidence intervals?

Its own page says so: “Siftly reports a mention rate per model with confidence intervals, so a drop is a signal rather than a coin flip.” The page does not state the interval method or confidence level, so treat it as a published claim rather than an audited one.

Is the Siftly free plan useful?

For one specific job, yes. Ten prompts on two engines refreshed weekly will not measure a rate, but it will tell you whether your category produces AI answers with citations at all, and that question decides whether any of this work can succeed for you.

Is Siftly good value?

Per answer collected it is the cheapest in our comparison, $0.0055 on Scale against $0.021 for Rankscale and $0.082 for Athena HQ. It buys that by selling depth rather than breadth: 150 prompts run often, not a thousand prompts run once. Whether that is value depends on whether your category fits in 150 questions.