By skyforbes 
Does ChatGPT Plus ($20 subscription) still outperform Claude and Perplexity?
During my last client project, I promised myself I’d finally find out whether my $20 ChatGPT Plus subscription is still worth it. I use ChatGPT mostly for research, editing, and writing, so that’s what I was going to test. The comparison is not entirely fair — I’m running the paid version of ChatGPT against the free plans of Perplexity and Claude. Why? I have been using ChatGPT from the start, and I wouldn’t consider switching unless the free versions came close to or matched what I use now.
The project that prompted this was a 100-page, tightly researched course script on a health topic. The timeline was so short that I only accepted it because I was explicitly allowed to use AI in any way I saw fit, as long as I was transparent about it. Without that flexibility, meeting the deadline with the required quality would have been impossible. I gave the client a discount (possibly more than I should have) and then set to work, collaborating with — and occasionally cursing at — ChatGPT.
Here’s what it did well overall:
- Outline and structure
- Formatting and editing
- Extracting information, summarizing, and comparing uploaded papers
Here’s where it still falls short:
- Literature search: Hallucinations remain common. It comes up with non-existent papers, provides broken links, or confidently cites non-existent sources.
- Truthfulness: Every claim still needs to be checked manually. The errors can be obvious but also subtle. Facts that sound plausible are sometimes fabricated.
That’s why I wanted to see whether Claude or Perplexity could at least match these results on their free plans. If they could, I’d still have reasons to complain — but at least I wouldn’t be paying $20 a month for the privilege. I didn’t include the free ChatGPT version in this test, as I already know it wouldn’t match the paid one. I have heard a lot about the research capabilities of Claude and Perplexity’s free tiers but never used them, so they make more sense as benchmarks.
Here are the three prompts I used for the comparison — they are simplified versions of the kinds of prompts useful for literature reviews, as they cover the three steps of literature search, extraction/synthesis, and academic rewriting.
1. Literature Search Prompt
Find 5 recent (past 5 years) peer-reviewed studies on obsessive-compulsive disorder (OCD) that would be most relevant for a literature review on the topic. Focus on research that is methodologically solid and clinically meaningful. Provide the references in full, formatted in APA 7 style, including authors, year, title, journal, and DOI if available. Include a DOI only if it resolves to a valid landing page. If no valid DOI exists, give the publisher URL or PubMed URL instead. Do not invent or guess DOIs.
2. Extraction + Summarization / Synthesis Prompt
For one of the studies you just listed, extract and summarize the following in detail:
— Study aims
— Methodology (participants, measures, procedures, analyses)
— Key results and conclusions
Then, add a short discussion paragraph that connects these findings to broader trends in OCD research (e.g., diagnosis, psychotherapy or treatment with SSRIs, comorbidity, or underlying mechanisms). Provide exact quotes with page numbers when possible and necessary. Fact-check your text before submitting.
3. Writing / Rewriting Prompt
Rewrite the summary and discussion from above in clear, fluent academic English suitable for inclusion in university or college course materials. Maintain the meaning and structure but improve coherence, conciseness, and readability.
RESULTS
Prompt 1: Literature Search
For the first prompt, I asked each tool to find five recent, peer-reviewed studies on OCD that would be suitable for a literature review, with correct APA-formatted references and valid DOIs.
Claude came back fast with a well-structured list. Four out of five studies were correct, one citation, however, was wrong: mismatched authors and a fake DOI. Still, for a free tool, the overall quality was strong.
Perplexity performed best on this task. All five studies were correct and properly formatted, though one was from 2018. However, it immediately warned me that my query had been upgraded to “Pro Search,” with only two enhanced searches left for the day. This is important: the free plan wouldn’t sustain the volume of queries I typically need for my work.
ChatGPT Plus came last. Only two out of five references were correct; the others had wrong or non-existent DOIs. It also took the longest, about two and a half minutes. This matches my experience, ChatGPT has always struggled with literature retrieval.
Bottom line: Perplexity wins this round, Claude follows with one serious citation slip, and ChatGPT trails behind. However, Perplexity’s usage limit would make it impractical for most projects without upgrading.
Prompt 2: Extraction + Summarization / Synthesis
For this prompt I asked the AI to extract information (study aims, methodology, key result/conclusions and a short discussion) from one of the studies correctly listed.
Claude provided a reason for his choice and came up with an impressively detailed, comprehensive breakdown of the study aims, methodology and a thorough discussion. However, immediately after Claude’s answer I got a message that I had reached my session limit for the free plan and that it would reset 4.5 hours later. Depth is excellent, but hitting the free-tier-cap after two sessions would make the free plan unfeasible for me.
Perplexity came up with a short, correct and concise summary, with a brief discussion of clinical relevance. It provided much less methodological detail than Claude.
ChatGPT Plus produced a solid, mid-length summary that followed the requested structure and stayed true to the source. The summary was more detailed than the one provided by Perplexity, but less granular than Claude’s. In line with the other two, it didn’t provide page numbers for direct quotes.
Bottom line: There is no clear winner. Depending on the goal, I’d choose Claude for maximum depth, Perplexity for a quick, short overview, and ChatGPT Plus for a reliable middle ground. All three did their job of extracting information well, but none of them were able to supply the requested page numbers.
Prompt 3: Rewrite for Course Materials
For the third prompt I asked for a concise rewrite of the summary to make it suitable for the inclusion in course materials.
Claude (next session — I had to wait to continue to use the free plan) delivered an almost course-ready rewrite with strong topic sentences, clear sectioning, and crisp transitions. It preserved nuance while reducing redundancy. The waiting time on the free tier, however, would be a practical limitation for continuous editing.
Perplexity produced a polished, succinct rewrite. Very readable for slides or a short handout, but, depending on specifications, not detailed and comprehensive enough for instructors or advanced students.
ChatGPT Plus produced a concise academic rewrite suitable as instruction material. Its performance in this category was solid, as expected. As I have the plus plan, there were no session limits.
Bottom line: Again, no clear winner. For turning draft notes into usable course text, Claude, Perplexity and ChatGPT Plus all delivered dependable results. Claude produced the most nuanced text, Perplexity a quick, easy to skim summary with less methodological detail. ChatGPT Plus had the advantage of not having session limits because I have the paid plan.
How much (manual) rewriting is still needed to humanize the text depends on your standards, stylistic expectations, and available time. In my last project, I simply didn’t have the luxury to rewrite everything from scratch; the output had to be good enough with minimal polishing.
Overall summary
I ran this test to see if I could replace my ChatGPT Plus subscription with the free plans of Perplexity or Claude. The answer for now is a clear no, because free plans don’t offer the same capacity and have restrictions. My test has shown, that for research and writing on this level, a paid plan is still essential.
However, I am thinking of adding Perplexity and Claude to my toolkit to use strategically alongside ChatGPT Plus. Layered workflows may turn out to be more efficient and more cost-effective, than sticking to a single AI tool. Overall, I found Claude’s and Perplexity’s performance impressive: Perplexity stood out for quick, accurate literature searches and concise overviews, while Claude surprised me with its ability to produce exceptionally detailed academic breakdowns of individual studies.
ChatGPT’s results were as usual: solid in some areas, weak in others. I will continue to use it for now, as its familiarity is hard to replace. I am used to the daily back-and-forth with ChatGPT. It has become part of my workflow, not just for research, but for shaping and refining prompts and ideas as I go.
Overall, a strategic combination of all three tools could deliver the best results. A natural next step would be to compare the paid versions directly. Perplexity’s Pro plan claims to include access to both Claude and ChatGPT, which would be ideal, but the practical limits of that setup remain to be tested.
==> If you’ve tried different AI tools for your own research or writing, I’d love to hear how they fit into your daily work.