Psychology Papers and AI Detection: What APA 7 Recommends Before You Write

Large parts of a psychology paper follow guidance written before you sit down to draft. APA recommends verb tenses by section, JARS identifies information that supports transparent reporting, and the numbers guide sets formats for statistics. Your course brief or target journal decides which reporting items are required for this paper. Turnitin, in a separate document, says its classifiers learn word-probability patterns and that false positives can involve text with little structural variation. A flag in Method or Results does not by itself establish that AI wrote the section, and deleting study facts or required disclosures is not an honest fix.

One thing to be clear about before anything else: the Turnitin AI writing detection FAQ used here does not make a claim about APA. We read that page in full on 27 August 2026, 31,712 characters of rendered text, and searched it. `APA` appears 0 times as a word. `MLA` 0. `style guide` 0. `citation` 0. `discipline` 0. `psychology` 0. The nearest thing on the page is a line about training data covering "less common subject areas such as anthropology, geology, sociology, and others."

The instrument was working: on that same page `qualifying text` returns 4 hits, `word probability` 2, `structural variation` 1. So the link drawn in this article is ours, assembled from two sets of documents that never mention each other. We have counted this page before with a different question in mind, and Why AI Detectors Flag Well-Written Essays reports what happened when we searched it for any vocabulary of writing quality.

If you learned to write a Method section by studying papers from the same subfield, you probably followed their shape as well as their content. That is one way conventions pass from published examples into a new paper.

Now add APA's tense recommendations and the reporting items your course or target journal expects. The participant paragraph may have little structural room left.

If English is not your first language the pull is stronger, because a model sentence carries the grammar along with the format. Whether that shows up in a score is a separate question with its own vendor material: Non-Native Speaker Writing Flagged as AI? What the Documentation Says.

Start with tense, because most people think of APA as a citation format and stop there. APA's Verb Tense page gives a table of recommended tenses by paper section, with its own examples:

Read the right-hand column again. Those are the exact sentence openings people point at when they say a paper "sounds like AI." They are printed on APA's own website, and the page notes the guidance sits in Publication Manual section 4.12.

The reporting guidance covers content as well as grammar. JARS-Quant Table 1 is titled "Information Recommended for Inclusion in Manuscripts That Report New Data Collections Regardless of Research Design," and it reads like a packing list:

• "Report inclusion and exclusion criteria, including any restrictions based on demographic characteristics."
• "Report major demographic characteristics (e.g., age, sex, ethnicity, socioeconomic status)."
• "percentage of sample approached that actually participated", plus how the intended and achieved sample sizes were determined.
• "results of all inferential tests conducted, including exact p values if null hypothesis significance testing (NHST) methods were used".
• "effect-size estimates and confidence intervals on estimates that correspond to each inferential test conducted, when possible".

That title matters. APA's own overview of JARS calls it a "structured guide," says educators and students can use it, and tells authors that journals communicate their use in the journal's instructions. Adopting APA 7 does not, by itself, turn every line of Table 1 into a universal requirement for every student paper or journal manuscript.

Then the digits. APA's Numbers and Statistics Guide says to report exact p values to two or three decimals, to write anything below .001 as "p < .001", and, in text, to "Use the term, not the symbol, for statistics" with "the means were" as its example.

Put those three documents on the desk together and a Method paragraph may have little room left. The tense is guided, the reporting items are mapped, and the digits are formatted. How much of that map is mandatory still depends on the assignment, research design, and target journal.

Turnitin's FAQ describes the difference it is looking for in statistical terms:

"Our classifiers are trained to detect these differences in word probability and are adept at the particular word probability sequences of human writers."

The same page also says the model "is not explicitly programmed to evaluate specific signals such as 'burstiness,' 'perplexity,' or other individual metrics sometimes referenced in public discussions." Read them together and the denial is aimed at two metrics that circulate in public discussion, not at the statistics of word choice itself. Word Probability and Turnitin: What the Classifier Actually Reads takes that apart properly.

The same FAQ then lists what its own false positives tend to look like:

"Sometimes false positives (incorrectly flagging human-written text as AI-generated), can include content without a lot of structural variation, text that literally repeats itself, or text that has been paraphrased without developing new ideas."

The sentence after that one is the one worth carrying into a meeting:

"If our indicator shows a higher amount of AI writing in such text, we advise you to take that into consideration when looking at the percentage indicated."

An APA Method section can have little structural variation because style guidance and reporting expectations narrow some choices. That resemblance is a possible connection, not proof that APA formatting caused a score. Nothing in it settles whether a particular score is right or wrong for a particular paper.

The number on the report is computed from part of your document, not all of it, because Turnitin defines what it reads:

"This qualifying text includes only prose sentences, meaning that we only analyze blocks of text that are written in standard grammatical sentences and do not include other types of writing such as lists, bullet points (short non-sentence structures), or other non-sentence structures."

That definition lands unevenly across an APA paper, and psychology is a bad case, because the manual tells you not to restate in prose the statistics that already sit in a table.

In a well-cited psychology paper the reference list can run several pages, and the tables can hold most of the numbers you generated. Both sit outside or largely outside the thing being scored, while your Method prose sits fully inside it. What Is "Qualifying Text" in Turnitin AI Detection? covers the general rule, and Does Turnitin Read Text Inside Tables? covers the table case.

This is the distinction that actually changes what you do next, and it is the one most revision advice skips.

The three sources do different jobs. APA recommends tenses, JARS offers reporting standards for the relevant research design, and the numbers guide gives formats for statistics. Your course or journal instructions decide which reporting items are mandatory. None of those sources writes your sentences for you. "The results indicate" and "Future research should explore" are illustrations of tense, not required phrasing.

So the paper splits in two.

Facts, and any required disclosures, stay fixed. If your study used inclusion and exclusion criteria, report them accurately where the assignment or journal requires them. The same applies to achieved sample size and its determination, exact p values, effect sizes, confidence intervals, and reliability coefficients. You can change how a sentence is expressed; you cannot change the underlying study or delete a required item to make a paragraph look less templated. Methods and results: what you can rephrase and what must stay exact goes through the line between expression and scientific state.

Your explanation. Why the question matters. What the gap in prior work is. What this result means given how the measures behaved. JARS recommends interpreting results "taking into account sources of potential bias and threats to internal and statistical validity" and discussing generalizability. It does not tell you what to conclude. If your Discussion reads as interchangeable with every other Discussion, that is habit rather than a sentence supplied by the manual.

1. Look at the highlighting before the number. A methods chapter that arrives as one continuous highlighted block has a mechanical explanation worth understanding first: Turnitin flagged my whole methodology section.
2. Count the prose, not the pages. If the analysed prose in a short report is only a few hundred words, the vendor says prediction becomes "mostly 'all or nothing'". See Short Documents and the All-or-Nothing Problem in Turnitin.
3. Mark the requirements before you touch any wording. Put the assignment brief or journal instructions beside the relevant JARS table. Keep every item your target actually requires, and keep every study fact accurate. JARS is the reporting map, not proof that every item is mandatory in every APA paper.
4. Bring the vendor's own position. Its FAQ says the percentage "should not be used as the sole basis for action or a definitive grading measure by instructors." What your instructor is told to do when your AI percentage is high collects that guidance.
5. If someone quotes a threshold at you, Is 20% AI too high? explains why no published number works as a pass mark.

Rewriting an APA Results paragraph by hand has one specific failure mode. You are shifting sentences around numbers that carry italic rules, decimal rules and a rule about not repeating anything that already appears in a table. The thing at risk is the numbers and the formatting, not the prose.

The narrow reason HumanPen belongs in this article is the approval screen, not a promised score. We make HumanPen, so read this as first-party information. After you select the Results prose you are willing to revise, the tool shows the paragraph-level scope before the job starts. Use that screen to check that no table note, numeric statement or neighboring paragraph slipped into the target. You can also import an existing Turnitin or iThenticate report to locate candidate passages. The returned document remains editable and is designed to preserve terminology, citations, styles and structure; still compare every statistic with the analysis output. Eligible passages can be processed again at no extra charge. Rewriting only the paragraphs a Turnitin report flagged covers the report-driven route.

What the score will be afterwards is not something anyone can promise you in advance. Whatever you use, put the returned statistics side by side with your analysis output before you submit. A p value that lost a decimal is a worse problem than a highlighted sentence.

Does using APA format itself raise an AI score? The AI detection FAQ read for this article makes no such claim and does not mention APA, MLA or style guides. It does say false positives can involve text with little structural variation. APA guidance and reporting expectations may narrow variation in some sections, but that does not establish that APA formatting caused a score.

Should I rewrite my Method section so it sounds less formulaic? Not by changing study facts or removing an item your course or journal requires. Use the relevant reporting standard as a checklist, confirm what applies to this paper, then revise expression rather than scientific content.

Do my tables and reference list count toward the percentage? Only qualifying prose sentences are analysed, and a 2023 release note says bibliographies are excluded when the AI writing report is processed, with resubmission needed for that change to apply to an older report. So the percentage describes part of your document, not all of it.

My entire Results section is highlighted and I ran the analysis myself. Running the analysis establishes where the numbers came from; it does not by itself establish who wrote the prose. If you wrote the Results text and it has little structural variation, it may fit one feature in the vendor's false-positive description. The percentage still is not a sole basis for action. Take the report, data files and drafts to the conversation rather than treating the highlight as a verdict.

Does this apply to APA 6 papers, or to qualitative psychology? The tense guidance and reporting standards above are the seventh edition and JARS-Quant. Qualitative work has its own APA standards. In either case, check the relevant standard and the requirements of your course or target journal before deciding what can change.