You just published a review of a client's new pricing page. Four items, two screenshots, a few claims you typed while looking at the screen. One of them says their annual plan is priced above the two competitors you named. You are fairly sure. Fairly sure is not something you want to put in front of a paying client.
Here is a way to catch that before you hit send: take the review's markdown, paste it into Perplexity, and ask it to check the factual claims against sources. You are not asking Perplexity to write the review. You are using it as a second reader that can pull citations.
Why the markdown, not the link
Every published Cobalt Capture review has a plain-text version at /r/<slug>/markdown. That is the format built for machines to read, and it is what you want here. It has your item text, the pin numbers, and image references, with none of the page chrome that trips a model up. You could hand Perplexity the /r/<slug> link and it would fetch the markdown automatically, but pasting the text gives you a clean, visible record of exactly what you asked about. When you want the whole workflow spelled out, the Perplexity workflow page covers fetching by link and by paste.
The actual input
This is the markdown copied straight from the review, trimmed to the item that matters:
## Item 2 - Pricing table (pin 1)
The annual plan lists at $228/yr. Pin 1 marks the price cell. This is higher than both Acme ($180/yr) and Baxter ($204/yr), so "most affordable in the category" in the hero is a stretch.
Then the prompt, typed into Perplexity:
Below is a product review in markdown. Check each factual claim in Item 2 against current public sources. Tell me which claims you can confirm, which you cannot, and cite what you find. Do not rewrite the review.
The actual result
Perplexity came back with three lines that changed what got sent:
- It confirmed Acme's annual plan at $180 and linked the current pricing page.
- It could not confirm Baxter at $204. The number it found was $216, with a citation, and it flagged that Baxter had changed pricing recently.
- It noted the $228 figure was the client's own page, which it could not independently verify, and told you to trust your screenshot for that one.
So the claim "higher than both" survived. The specific number for Baxter did not. You go back into the review, fix $204 to $216, and now the item holds up if the client checks. That is the whole point: a wrong competitor price in a review you sent is the kind of small miss that makes a client doubt the rest of your work.
Fix the item, republish, send
Because the review is still editable, you open it, correct the Baxter figure in Item 2, and republish. The link stays the same. Nothing about the client's experience changes except the number is now right. This only works while the review is inside its editable window: after seven days a review goes read-only unless you are on the $5 Pro plan, which keeps every review editable and reopens read-only ones on the spot. If you edit reviews days after publishing, that is the reason to be on it.
The shared link keeps working for the client either way, so a read-only review is not lost. You just cannot correct a fact in it without reopening, which Pro does instantly.
Where this fits, and where it does not
Perplexity is checking claims, not judging design. It will not tell you a button is too small or a flow is confusing. Those are the things you saw with your own eyes and marked with a pin, and no second reader replaces that. Use Perplexity for the checkable claims: prices, dates, competitor names, version numbers, anything a client could look up and find wrong. Keep it away from the subjective calls.
The same move works well on a competitive analysis, where half the review is claims about other products, and on a pricing page teardown, where a single wrong figure undercuts the argument. For those, structuring the review so the facts sit in their own items makes the check faster. A short, clear item like the one above is easy for a model to isolate; a paragraph mixing three claims and an opinion is not. A worked example of structured feedback for an LLM shows how to split them.
Start the review, then check it
The workflow assumes you already have a published review with a markdown URL. If you do not, start a review in a browser tab, capture the pricing page, pin the price cell, type your claim, and publish. Copy the markdown, paste it into Perplexity with the prompt above, and fix whatever comes back wrong. The client gets a review where every number holds, and you spent two minutes to get there.