Poster Presentations and AI Detection: Why Scores Get Unreliable

Posters are not written the way essays or journal articles are. They rely on bullet points, short phrases, and visual elements. Turnitin's documentation says the model only analyzes qualifying prose sentences and does not reliably detect AI in non-prose formats. When a poster is submitted as a document, the percentage may reflect a small slice of prose while the highlights show almost nothing. The disparity between the two is a known behavior, not a glitch.

The FAQ is clear about what counts as qualifying text:

"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."

The FAQ also states:

"The model does not reliably detect AI-generated text in the form of non-prose, or code, nor does it detect short-form/unconventional writing such as bullet points (short non-sentence structures)."

The next sentence explains the consequence:

"This means that a document containing several different writing types would result in a disparity between the percentage and the highlights."

For a poster, this disparity is extreme. Most of the text on a poster is in bullet points or short labels. If you submit a poster as a document and get a high AI percentage, that percentage may be based on just a few qualifying prose paragraphs, while the bullet points that make up the bulk of the poster are invisible to the detector.

One Turnitin help article says the model's limitations extend further than the FAQ's list:

"The model does not reliably detect AI-generated text in the form of non-prose, such as poetry, scripts, or code, nor does it detect short-form/unconventional writing such as bullet points, tables, or annotated bibliographies."

The next sentence from that article:

"This means that a document containing several different writing types would result in a disparity between the percentage and the highlights."

Posters frequently use tables for results, comparisons, and timelines. If tables are listed as not reliably detected, then a significant portion of a poster's content falls outside the model's reliable coverage. The percentage you see may be computed from the few prose paragraphs sandwiched between tables and bullet lists.

There is an important update from August 2023. A release note says:

"We are now able to process long-form prose text in tables."

The next sentence is:

"Resubmit to reprocess existing submissions that contain tables."

This means tables with long-form prose sentences (full grammatical sentences inside table cells) are now processed. Short labels, numbers, and data entries inside tables are still non-prose and remain outside the model's analysis. For posters, this is a partial improvement. A results table with full sentences in each cell would now be analyzed. A table with bare numbers and short labels would not.

The FAQ describes how submissions are processed:

"When a paper is submitted to Turnitin, sentences from the submission are extracted and segmented into overlapping sections for prediction analysis. Each segment is classified by the AI detection model and given a value between 0 and 1, denoting the probability of the text being likely human or AI-generated. Each qualifying sentence within these segments inherits the segment's score."

Only qualifying prose sentences are extracted. The aggregation then produces the document-level score:

"These sentence scores are further aggregated and used to compute the overall document AI writing score."

For a poster, the extraction step may pull very few sentences. The few paragraphs of prose on a poster might be the abstract, a short introduction, and a concluding remark. Everything else is bullet points and table entries. The score is computed from that small set of prose, which means it reflects a narrow sample. How to read a Turnitin AI writing report starts from the same gap between the number and the page.

Posters submitted as documents are short. The FAQ warns about this:

"In shorter documents where there are only a few hundred words, the prediction will be mostly 'all or nothing' because we're predicting on a single segment without the opportunity to overlap."

The next sentence:

"This means that some text that is a mix of AI-generated and original content could be flagged as entirely AI-generated."

A poster with only 300 or 400 words of qualifying prose falls squarely in this range. With so few segments, there is no averaging effect. One segment that scores high can label the entire document as AI-generated. The FAQ's own next sentence about the disparity between percentage and highlights becomes very visible here. You might see a high percentage but almost no highlighted text on the poster itself, which is one of the cases in why your AI writing score is missing, inconsistent, or will not download.

The FAQ describes what tends to trigger false positives:

"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."

Poster prose tends to be repetitive. The same key findings appear in the abstract, the results section, and the conclusion. The language is compressed and uniform because space is limited, and uniformity is close to the thing what AI detectors measure says the classifier is picking up. The FAQ continues:

"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."

For posters, this advice is doubly relevant. The prose is short, repetitive, and structurally uniform. All three factors compound the risk of a misleading score.

If your poster presentation received a Turnitin AI score, here is what to keep in mind:

• Bullet points and short labels are not qualifying text. The score does not reflect them.
• One Turnitin help article says tables are also not reliably detected, though long-form prose in tables has been processed since August 2023.
• The disparity between the percentage and the highlights is expected behavior for mixed-format documents.
• Short qualifying prose triggers "all or nothing" predictions with minimal segment overlap.
• Poster prose is repetitive and uniform, matching the false positive profile the FAQ describes.
• A high percentage with few highlights should be treated with caution, not as a definitive verdict.

If you want to address the flagged passages, import your Turnitin report and work through them. Eligible passages can be re-run at no charge.