Skip to content
October 2026 results: WriteHuman holds #1, StealthGPT jumps to #2. Read the analysis →
HumanizerBench

Head to head · October 2026 cycle

Phrasly vs Walter Writes

Walter Writes finished 1.7 points ahead of Phrasly in the October 2026 cycle, 63.53 to 61.79, ranking #10 against #12 in a field of 14. Walter Writes scored higher on 0 of 6 detectors and 5 of 6 writing categories. Phrasly came out ahead on Pangram (85.6 vs 0.0), ZeroGPT (94.3 vs 71.6) and bypass rate (95.3 vs 72.8).

61.79 /100
overall score
#12 /14
rank ▼ 4
+1.74
Walter Writes leads →
Detectors
6 – 0
Categories
1 – 5
Components
2 – 3
63.53 /100
overall score
#10 /14
rank -

30 prompts · Last tested · Methodology v1.3.0

Score components

What the overall score is made of: bypass rate weighs 42%, meaning preservation 32%, readability 16%, consistency across categories 10%. Output-quality penalties come off the total. How scoring works →

Bypass rate Phrasly by 22.6
95.3 72.8
Meaning preservation Walter Writes by 20.5
49.9 70.4
Readability Walter Writes by 19.3
44.2 63.5
Consistency Phrasly by 4.2
96.9 92.7
Penalties fewer is better Walter Writes by 2.0
−11.0 −9.0

Detector by detector

Share of each tool's outputs that a detector classified as human-written. Phrasly took 6 of 6.

GPTZero Phrasly by 10.8
98.6 87.8
Originality.ai Phrasly by 5.1
96.6 91.5
Copyleaks Phrasly by 10.7
100.0 89.3
Winston AI Phrasly by 0.5
97.0 96.5
ZeroGPT Phrasly by 22.7
94.3 71.6
Pangram Phrasly by 85.6
85.6 0.0

Writing categories

Category score per writing context: bypass credit only for real rewrites, blended with the category's own meaning preservation and readability. Small categories swing; the prompt count is shown on each row. How category scores work →

Academic Essay Application Essay Blog Post Marketing Copy Discussion Board News Article Academic Essay: Phrasly 63.7 · Walter Writes 71.4 Application Essay: Phrasly 63.6 · Walter Writes 66.1 Blog Post: Phrasly 57.8 · Walter Writes 61.2 Marketing Copy: Phrasly 45.8 · Walter Writes 53.2 Discussion Board: Phrasly 52.9 · Walter Writes 37.9 News Article: Phrasly 16.8 · Walter Writes 56.0
Phrasly Walter Writes Outer ring = 100
Academic Essay 6 prompts Walter Writes by 7.8
63.7 71.4
Application Essay 6 prompts Walter Writes by 2.5
63.6 66.1
Blog Post 6 prompts Walter Writes by 3.5
57.8 61.2
Marketing Copy 6 prompts Walter Writes by 7.5
45.8 53.2
Discussion Board 3 prompts Phrasly by 15.0
52.9 37.9
News Article 3 prompts Walter Writes by 39.3
16.8 56.0

Head-to-head history

Overall score in every cycle both tools were tested. Phrasly has finished ahead in 3 of 5 shared cycles.

50 60 70 80 June 2026 July 2026 August 2026 September 2026 October 2026 June 2026: Walter Writes 67.24 (#10) June 2026: Phrasly 71.10 (#8) July 2026: Walter Writes 62.84 (#7) July 2026: Phrasly 61.47 (#9) August 2026: Walter Writes 60.83 (#9) August 2026: Phrasly 63.30 (#7) September 2026: Walter Writes 61.70 (#10) September 2026: Phrasly 67.58 (#8) October 2026: Walter Writes 63.53 (#10) October 2026: Phrasly 61.79 (#12) 61.8 63.5
Phrasly Walter Writes
Cycle Phrasly Walter Writes Margin
October 2026 61.79 #12 63.53 #10 → 1.7
September 2026 67.58 #8 61.70 #10 ← 5.9
August 2026 63.30 #7 60.83 #9 ← 2.5
July 2026 61.47 #9 62.84 #7 → 1.4
June 2026 71.10 #8 67.24 #10 ← 3.9

Where each one wins

Every measure above that one tool won outright, largest margin first within each group.

Phrasly

9 wins
  • Bypass rate Score 95.3 vs 72.8
  • Consistency Score 96.9 vs 92.7
  • Pangram Detector 85.6 vs 0.0
  • ZeroGPT Detector 94.3 vs 71.6
  • GPTZero Detector 98.6 vs 87.8
  • Copyleaks Detector 100.0 vs 89.3
  • Originality.ai Detector 96.6 vs 91.5
  • Winston AI Detector 97.0 vs 96.5
  • Discussion Board Category 52.9 vs 37.9

Walter Writes

8 wins
  • Meaning preservation Score 70.4 vs 49.9
  • Readability Score 63.5 vs 44.2
  • Penalties Score −9.0 vs −11.0
  • News Article Category 56.0 vs 16.8
  • Academic Essay Category 71.4 vs 63.7
  • Marketing Copy Category 53.2 vs 45.8
  • Blog Post Category 61.2 vs 57.8
  • Application Essay Category 66.1 vs 63.6

Same prompt, both outputs

Both tools were given this AI-written passage in the October 2026 cycle. Below it is what each one returned. This prompt was picked by a fixed rule, not by how the tools scored.

News Article · News Article · written by claude-sonnet-5-5 · 444 words

**Digital Privacy Laws Struggle to Keep Pace With Data-Hungry Technology**

As lawmakers worldwide celebrate landmark privacy regulations, researchers and advocates warn that the laws are straining against the realities of modern data collection, leaving significant gaps in how personal information is protected.

Europe's General Data Protection Regulation, California's Consumer Privacy Act, and a growing patchwork of state and national laws have given individuals new rights to access, delete, and limit the use of their data. Yet critics say these protections often prove more symbolic than practical.

One central problem is consent. Most privacy frameworks rely on the idea that users agree to data practices after reading disclosures. In reality, few people read lengthy privacy policies, and fewer still understand them. A frequently cited study estimated that reading every policy an average person encounters in a year would take weeks of working time. Consent banners, meanwhile, are often designed to nudge users toward clicking "accept."

Enforcement presents another obstacle. Regulators are chronically understaffed relative to the thousands of companies they oversee. Investigations can take years, and fines, while sometimes large in absolute terms, may amount to a small fraction of revenue for major technology firms. Some companies treat penalties as a cost of doing business.

The borders of the internet also complicate matters. Data flows across jurisdictions in milliseconds, but legal authority stops at national lines. Disputes over transferring European data to the United States have repeatedly invalidated agreements, creating uncertainty for businesses and users alike.

Technology itself is outrunning legal definitions. Privacy laws typically protect "personal information," but advances in data analysis allow companies to re-identify supposedly anonymous records by combining datasets. Inferences drawn by algorithms, such as predictions about health, finances, or political leanings, occupy murky legal territory because they are generated rather than provided by the user.

Meanwhile, large categories of data collection fall outside existing rules. Data brokers assemble detailed profiles from public records, app activity, and purchased information, often with little direct relationship to the people profiled. Government access to commercially held data raises further questions, as agencies can sometimes buy information they would otherwise need a warrant to obtain.

Privacy advocates argue for a shift away from individual consent toward limits on what companies may collect and do in the first place, an approach known as data minimization. Some proposals would ban certain practices outright, such as selling sensitive location data.

Industry groups counter that overly rigid rules could stifle innovation and burden small businesses. The debate is likely to intensify as artificial intelligence expands the appetite for training data.

For now, experts say, individuals remain largely responsible for protecting themselves, a burden that few are equipped to carry.

Show the full input

438 words (input 444)

As governments continue to develop more strict privacy laws researchers are beginning to show concern over whether these laws can effectively cover the amount of data being collected by companies. Many countries have developed privacy acts that allow their citizens to know what data companies are collecting and can delete or stop companies from using their data. Although there are flaws with these new regulations many people believe it to be a positive. There are many drawbacks to these laws. Privacy policies allow companies to gain your consent to use your data. Since most people do not read privacy policies its almost impossible to truly tell if someone knows their data is being used. It's been said that it would take the average person 7+ hours a day for one year just to read all the policies you'd come across. There are many reasons why people don't take the time to read these policies and some people even say the companies purposely do this to ensure consent. Another issue with privacy laws is enforcement. There aren't enough employees within regulatory agencies to keep up with all of these companies. When cases are investigated it can take multiple years. Once fines are given to companies it can sometimes seem cheap compared to their earnings. Because of how global our world has become with internet and data. There is so much gray area when it comes to international laws and regulations. Recently we've seen many examples of why moving EU data to the US isn't possible and will create issues. As our technology advances there is more gray area around privacy laws. What is considered personally identifiable information? Companies can use databases of information and figure out who people are by matching different sets of data together. Also with all the new information being derived from peoples data where does the line get drawn when you didn't give that information. Not all forms of collecting data will fall under these new laws. Companies such as data brokers buy and sell peoples information with no knowledge of that person. Governments also can use this data to find information about people that they wouldn't have been able to obtain without a warrant.

With more technology and companies developing ways to profit off peoples data. There are calls for change from those who believe that people should have more control over their information. One example would be limiting what information a company can collect on you.

Privacy laws are something that will continue to grow and evolve with our world. Currently its up to each of us to help secure our own information.

Show the full output
  • GPTZero1.00passed
  • Pangram1.00passed
  • Winston AI1.00passed
  • ZeroGPT0.98passed
  • Copyleaks1.00passed
  • Originality.ai1.00passed

Passed 6 of 6 detectors.

Every Phrasly run in October 2026 →

762 words (input 444)

**Laws Protecting Digital Privacies Are Slowly Being Outpaced By The Amount Of Data That Companies Are Collecting From Technology Users**

While legislators around the world have been celebrating the passage of landmark digital privacy legislation, privacy advocates and researchers are warning that those laws cannot keep pace with the way we currently gather and use our data. As a result, there are many areas where the law does not provide adequate protection for users' personally identifiable information (PII).

Recent state and federal laws give consumers the right to review, delete, and restrict companies from using consumer PII. However, most critics of these laws believe that the protections being offered by these laws are more symbolic than actual.

A primary challenge facing all forms of digital privacy regulation is the issue of informed consent. While nearly all forms of online privacy regulation attempt to establish a connection between a company collecting consumer PII and obtaining the consumer's permission to do so, it is well established that most consumers do not actually read or comprehend privacy policies. One widely-cited study found that if one were to spend an average of just five minutes per day reading each and every privacy statement a typical Internet user encounters during the course of a single calendar year, he/she/they would be required to spend approximately two full months of work time solely to accomplish this task. Furthermore, even when consumers are presented with a consent banner, most simply click "yes" in order to continue accessing the service.

Another barrier to enforcing meaningful digital privacy regulation is the fact that regulatory bodies are grossly underfunded compared to the number of entities they regulate. Regulatory investigations into alleged violations of digital privacy law can take multiple years to complete. Even then, fines issued to violators tend to be based upon a percentage of revenue that may not be material enough to impact a major technology firm's bottom line. Additionally, some of the largest technology firms view fines as merely a normal operating expense.

Finally, the international nature of the Internet creates additional obstacles to enacting effective digital privacy regulation. Because data travels internationally at speeds measured in milliseconds, regulatory jurisdiction stops at national boundaries. For example, numerous times disputes regarding the transfer of personal data from Europe to the U.S. have resulted in invalidation of previously negotiated agreements and created uncertainty for both businesses and end-users.

In addition to regulatory challenges, the rapid technological advancements occurring in the area of big data analytics and artificial intelligence are rendering outdated much of the language contained within current digital privacy laws. Current laws generally define "personally identifiable information" (PII) as any information that could reasonably be used to identify an individual. However, recent advancements in big data analytics have allowed companies to create profiles of individuals who had previously been anonymized by linking together various datasets. Moreover, inferences made by algorithms concerning an individual's health, financial status, politics etc., exist in a gray area of law as they represent something derived from a combination of publicly available sources, app usage patterns, credit reports etc. versus data provided directly by an individual.

Additionally, several types of data collection activities are taking place which are beyond the scope of current digital privacy regulations. For instance, data brokerages are compiling comprehensive dossiers containing information on virtually every aspect of an individual's life through a variety of means including; public records searches, app usage data, information purchases etc. without having any direct relationship to the individual(s) being profiled. Similarly, government acquisition of commercially-held data has raised serious concerns as governments are able to purchase information that would normally require a warrant to acquire.

Advocates of digital privacy advocate for a paradigmatic shift in the manner in which digital privacy is regulated away from requiring individuals to provide explicit consent prior to data collection towards limiting the type and quantity of data collected by companies prior to collection (data minimization). Advocates also propose banning certain types of data-related activities entirely (e.g. sale of sensitive location data).

On the other hand, industry representatives argue that overly restrictive data collection limitations will hinder future innovations in technology and disproportionately harm small businesses due to compliance costs associated with implementing complex regulatory requirements. This battle is expected to escalate as AI continues to drive ever-increasing demands for high volumes of quality training data.

Currently, experts agree that until legislative bodies create and implement better digital privacy laws, individuals will bear the responsibility for protecting their own digital information -- a burden that few possess sufficient knowledge or resources to assume.

Show the full output
  • GPTZero0.99passed
  • Pangram0.00caught
  • Winston AI0.99passed
  • ZeroGPT0.94passed
  • Copyleaks1.00passed
  • Originality.ai1.00passed

Passed 5 of 6 detectors.

Every Walter Writes run in October 2026 →

Each chip is one detector's score for that output, from 0 (AI) to 1 (human). A score of 0.50 or higher counts as a pass and shows green; below that the detector caught it and the chip shows red.

Questions people ask

Is Phrasly better than Walter Writes?

Walter Writes finished 1.7 points ahead of Phrasly in the October 2026 cycle, 63.53 to 61.79, ranking #10 against #12 in a field of 14. Walter Writes scored higher on 0 of 6 detectors and 5 of 6 writing categories. Both tools ran the same 30 prompts, scored by the same 6 detectors, under methodology v1.3.0; every input, output and verdict is published.

Which bypasses AI detectors better, Phrasly or Walter Writes?

Phrasly posted the higher bypass rate, 95.3 vs 72.8, and scored higher on 6 of 6 detectors. Phrasly led on GPTZero, Originality.ai, Copyleaks, Winston AI, ZeroGPT and Pangram.

Do Phrasly and Walter Writes keep the original meaning?

Walter Writes preserved meaning better, 70.4 vs 49.9 on our 0–100 similarity scale. On readability, Walter Writes rated higher, 63.5 vs 44.2. Penalties this cycle: Phrasly −11.0, Walter Writes −9.0.