Felo vs Perplexity: Two Cited-Answer Tools, Different Friction

Both Felo and Perplexity sell the same promise: answers with citations instead of confidence. I scored Felo 4/5 after an anonymous multilingual session, and Perplexity 4/5 for a cited research synthesis I could rebuild into a table. The tie breaks on the job: open Felo for a fast factual trap check across languages without an account; open Perplexity when you need numbers, caveats, and a table you can reuse.

I tested them on different days with different questions, so this is not a same-prompt shootout. What I can compare is what each free path let me finish, and where each one stopped short of a five.

A three-part query with a lie planted in it

For Felo I stayed logged out on purpose. The query bundled a capital, a UNESCO listing, and a deliberate false premise: the capital of Argentina, whether Iguazu is a World Heritage Site, and whether Mount Aconcagua is in Chile.

About 21 seconds later the lead paragraph got all three right. Buenos Aires as capital. Iguazu inscribed in 1984 on the Argentine side and 1986 on the Brazilian side. Aconcagua came back with a bold not — Mendoza province, roughly 15 km from the Chilean border, not Chile. The panel showed 30 Source & Thinking. Wikipedia Buenos Aires and Britannica Buenos Aires both opened cleanly. The UNESCO World Heritage Centre URL was genuine; Cloudflare threw an interstitial before the page appeared.

The refusal is the part I would keep. Plenty of tools will happily answer "where in Chile is Aconcagua" as if the question were sound.

The same answer rewritten in Spanish, still anonymous

Still anonymous, I asked Felo to rewrite the result in Spanish and keep the citations. About 31 seconds later I had a full Spanish answer with the reference markers intact, and the Chile trap was still refused in the new language. The source count climbed from 30 to 41.

That is the multilingual case in one move: not just translating a summary, but holding the citation structure and the fact-check across languages. The longer walkthrough is in my Felo review.

Where Felo stops: a captcha before creation

Anonymous search held up. Creation sat behind a login, so I tried the email signup path. The verification code arrived in about 55 seconds — in Junk, with a 60-second on-screen countdown. After I entered the code, the page threw a reCAPTCHA image grid: select all squares with crosswalks. I do not solve those in these sessions. Skip returned "Verification failed." No account. Clicking Generate AI Slide produced a login wall: Google, Apple, or email.

So the honest read is that the anonymous path is the product I actually used, and it worked. The moment I wanted slides, an image captcha stood in the way.

Pricing was readable while logged out. Free forever includes 200 bonus credits a day, unlimited high-speed search, 5 professional searches a day, and 3 file analyses a day. Pro runs $14.99 a month, or $12.50 a month billed annually at $149.99 a year.

Seven seconds on fasting and HbA1c

My Perplexity session ran separately, earlier, in an Incognito Free window. The UI said sessions are not saved and expire after 24 hours. No signup and no card for the queries I ran.

The first question was dense: what 2023–2024 systematic reviews and randomized trials found about intermittent fasting and HbA1c in adults with type 2 diabetes, with studies cited and limitations called out. First useful prose landed in about seven seconds with 24 sources, and it carried real numbers — a 2023 RCT meta-analysis SMD of −0.08 (95% CI −0.20 to 0.04; P=.19), INTERFAST-2 at −7.3 mmol/mol (~−0.6 percentage points) versus control +0.1, EARLY 5:2 meal-replacement −1.9 points at 16 weeks versus metformin −1.6 and empagliflozin −1.5. A soft "This question looks difficult" card with Upgrade to Pro sat over the answer. The answer still finished on Free. The full session is in my Perplexity review.

A table rebuilt in eight seconds

I asked it to turn the evidence into a comparison table across the 2023 meta-analysis, INTERFAST-2, EARLY, and a 2024 16:8/14:10 trial, and to flag bundled interventions. About eight seconds later, 30 sources deep, the table held the caveats: insulin-dose adjustment in INTERFAST-2; meal replacements, calorie guidance, contact, and education bundled into EARLY. That was the moment I stopped testing "can it write" and started trusting it for structure.

Thirty-six seconds, then Chinese

A second query — remote work and software-engineer productivity from 2023–2025 studies, with causal versus correlational evidence — took about 36 seconds and produced serious synthesis across 25 sources: Gibbs/Mengel/Siemroth (2023) full-remote IT setting with output/hour down about 8–19%; Bloom/Han/Liang (2024 Nature) hybrid RCT, two days a week at home, no significant performance damage and lower attrition; BLS industry evidence labeled correlational. The Nature citation opened the real 12 June 2024 paper.

The wobble was language. The prompt was English; the whole answer rendered in Chinese. The retrieval path still worked. The language router did not. Another Upgrade to Pro card and a "Pro Free preview of advanced search enabled" banner sat alongside it.

What the pricing pages said without an account

On annual Personal pricing, Perplexity listed Pro at $17 a month and Max at $167 a month. Free was referenced as the base tier ("Everything in Free"), but the modal showed no separate numeric Free allowance. Felo's logged-out page was more forthcoming with the free daily caps and the two Pro prices above.

Pick by the job, not the brand

If I need a quick geographic or factual trap check — something with a planted false premise — and I want it rewritten in another language without losing citations, I open Felo and stay anonymous. That path worked, including Spanish on the first try. Budget patience if you need slides: signup hit an image captcha for me.

If I need literature synthesis with numbers, caveats, and a table I can rebuild, I open Perplexity. A Free Incognito session delivered all of that without an account. Watch for soft Pro upsells, an unexpected language flip on a hard prompt, and a pricing modal that will not show you a Free meter.

Both earned a four. Pick the shape of answer you actually need, then open the citations before you trust the sentence they sit under.