pith. machine review for the scientific record. sign in

arxiv: 2603.04925 · v2 · submitted 2026-03-05 · 💻 cs.IR

Recognition: no theorem link

Detecting RAG Advertisements Across Advertising Styles

Sebastian Heineking , Wilhelm Pertsch , Ines Zelch , Janek Bevendorff , Benno Stein , Matthias Hagen , Martin Potthast
Authors on Pith no claims yet

Pith reviewed 2026-05-15 15:56 UTC · model grok-4.3

classification 💻 cs.IR
keywords RAGadvertisement detectionLLM responsesentity recognitionadvertising stylesevasion robustnessnative ads
0
0 comments X

The pith

Entity recognition models can locate ads exactly in LLM responses and remain robust when advertisers switch styles.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds a taxonomy of advertising styles for LLMs based on explicitness and type of appeal. It then simulates advertisers changing styles to evade detection and tests multiple ad-detection methods on the resulting responses. Models that treat ad detection as an entity recognition task prove highly accurate at both identifying responses that contain ads and pinpointing their exact location within the text. These models stay largely effective even after style changes. In contrast, lighter models such as random forests and SVMs work well on seen styles but degrade when advertisers vary their approach, underscoring the need for efficient yet robust detectors suitable for end-user devices.

Core claim

Models trained to perform entity recognition on LLM outputs can exactly locate advertisements within those responses. These models detect the presence of ads with high accuracy and remain largely robust when advertisers alter the explicitness or appeal type of their messages, while lighter classifiers prove brittle under the same style shifts.

What carries the argument

Entity recognition models that identify and mark the precise spans of advertisements inside LLM-generated text, evaluated against a taxonomy of styles defined by explicitness and appeal type.

If this is right

  • Precise location of ads enables targeted blocking or highlighting without removing entire responses.
  • Robustness to style variation reduces the frequency of retraining needed when advertisers adapt.
  • Lightweight models require additional safeguards to maintain performance against evasion.
  • Practical ad blocking on low-resource devices becomes feasible if efficiency improvements close the gap with heavier models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same entity-recognition framing could apply to detecting other forms of generated manipulation such as misleading citations or biased summaries.
  • Deployment on real traffic would likely surface novel evasion tactics that the current taxonomy does not cover.
  • Combining entity signals with relevance or consistency checks might further strengthen detection without increasing model size.

Load-bearing premise

The simulated advertising styles and evasion attempts in the evaluation accurately reflect the diversity of real-world advertiser behavior and the styles discussed in marketing literature.

What would settle it

Testing the trained models on a collection of real deployed RAG responses that contain native ads drawn from live systems would show whether the reported robustness holds outside the simulated styles.

Figures

Figures reproduced from arXiv: 2603.04925 by Benno Stein, Ines Zelch, Janek Bevendorff, Martin Potthast, Matthias Hagen, Sebastian Heineking, Wilhelm Pertsch.

Figure 1
Figure 1. Figure 1: Examples of generated native ads in RAG responses [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Example responses for different advertising prompts. The chat window shows a user query for last minute travel and [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Prompt to create covert advertisements with rational appeals. The placeholders are filled with the information [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Ad detection odds ratios (95 % CI). For each classifier and new test set, we compared the odds of detecting an ad in the [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Average overlap in false negatives. The heatmap [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Entity recognition odds ratios (95 % CI). For each [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
read the original abstract

Large language models (LLMs) enable a new form of advertising for retrieval-augmented generation (RAG) systems in which organic responses are blended with contextually relevant ads. The prospect of such "generated native ads" has sparked interest in whether they can be detected automatically. Existing datasets, however, do not reflect the diversity of advertising styles discussed in the marketing literature. In this paper, we (1) develop a taxonomy of advertising styles for LLMs, combining the style dimensions of explicitness and type of appeal, (2) simulate that advertisers may attempt to evade detection by changing their advertising style, and (3) evaluate a variety of ad-detection approaches with respect to their robustness under these changes. Expanding previous work on ad detection, we train models that use entity recognition to exactly locate an ad in an LLM response and find them to be both very effective at detecting responses with ads and largely robust to changes in the advertising style. Since ad blocking will be performed on low-resource end-user devices, we include lightweight models like random forests and SVMs in our evaluation. These models, however, are brittle under such changes, highlighting the need for further efficiency-oriented research for a practical approach to blocking of generated ads.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper develops a taxonomy of RAG advertising styles in LLMs by combining explicitness and appeal dimensions drawn from marketing literature, simulates style-based evasion attempts by advertisers, and evaluates multiple ad-detection methods. It reports that models using entity recognition to locate ads within LLM responses are highly effective at detection and largely robust to style changes, while lightweight models (random forests, SVMs) are brittle; the work emphasizes the need for efficient detectors suitable for low-resource devices.

Significance. If the simulation results hold under real advertiser behavior, the work supplies a concrete taxonomy and evaluation framework for detecting generated native ads in RAG outputs, which is timely given the rise of blended advertising in retrieval-augmented systems. The explicit focus on lightweight models and end-device constraints adds practical value beyond purely accuracy-oriented detection studies.

major comments (2)
  1. [Evaluation / Simulation sections] The central robustness claim (entity-recognition models remain effective under style changes) rests on simulations generated from the taxonomy; the manuscript provides no quantitative validation (e.g., similarity metrics or human ratings) that the generated examples match observed real-world RAG ad styles or actual evasion attempts, which directly affects transferability of the reported robustness.
  2. [Abstract] The abstract states clear effectiveness and robustness findings yet supplies no quantitative metrics, dataset sizes, error bars, or baseline comparisons; without these details in the summary, it is impossible to judge whether the empirical results actually support the strength of the claims.
minor comments (2)
  1. [Taxonomy section] The taxonomy diagram would benefit from explicit axis labels and example sentences for each cell to improve readability.
  2. [References] Several citations to marketing literature on advertising styles are referenced but not listed in the bibliography; please ensure all are included.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. The comments have helped us clarify the scope of our simulation-based evaluation and strengthen the presentation of results. We address each major comment below and indicate the revisions made to the manuscript.

read point-by-point responses

  1. Referee: [Evaluation / Simulation sections] The central robustness claim (entity-recognition models remain effective under style changes) rests on simulations generated from the taxonomy; the manuscript provides no quantitative validation (e.g., similarity metrics or human ratings) that the generated examples match observed real-world RAG ad styles or actual evasion attempts, which directly affects transferability of the reported robustness.

    Authors: We agree that the lack of quantitative validation (such as similarity metrics or human ratings) for the simulated examples is a genuine limitation that affects claims about transferability to real-world advertiser behavior. Our evaluation uses controlled simulations derived from the taxonomy precisely to isolate the impact of style changes in a reproducible way, as no public datasets of labeled RAG advertisements with style annotations currently exist. In the revised manuscript we have added an explicit Limitations subsection that discusses the simulation approach, its assumptions, and the need for future human validation studies. We have also expanded the appendix with additional qualitative examples of the generated styles to aid reader assessment. These changes clarify the boundaries of our robustness results without overstating generalizability. revision: partial

  2. Referee: [Abstract] The abstract states clear effectiveness and robustness findings yet supplies no quantitative metrics, dataset sizes, error bars, or baseline comparisons; without these details in the summary, it is impossible to judge whether the empirical results actually support the strength of the claims.

    Authors: We accept this criticism. The original abstract was written in a high-level style that omitted key empirical details. In the revised manuscript we have updated the abstract to include the size of the simulated dataset, the main performance metrics (F1-scores) for the entity-recognition detectors under both original and style-changed conditions, and direct comparisons to the lightweight baselines (random forests and SVMs) that demonstrate their relative brittleness. We have also noted that results are averaged over multiple runs with standard deviations. These additions allow readers to immediately assess the strength of the reported findings. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical pipeline is self-contained

full rationale

The paper constructs a taxonomy from external marketing literature, generates simulated RAG responses with ads, trains detection models (including entity-recognition-based ones) on that data, and measures effectiveness plus robustness to style changes on held-out simulations. No equation or claim reduces by construction to a fitted parameter renamed as a prediction, no self-citation supplies a uniqueness theorem or ansatz that the current work merely renames, and the central results are direct performance numbers on freshly generated test sets rather than tautological re-derivations of inputs. This is standard empirical ML evaluation and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that the new taxonomy captures relevant marketing styles and that simulated evasions are representative; no free parameters or invented entities are introduced.

axioms (1)
  • domain assumption The taxonomy combining explicitness and type of appeal adequately covers the diversity of advertising styles from marketing literature.
    The paper states it combines these dimensions to build the taxonomy for LLM ads.

pith-pipeline@v0.9.0 · 5530 in / 1258 out tokens · 84802 ms · 2026-05-15T15:56:04.299115+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages · 3 internal anchors

  1. [1]

    Anu Adegbola. 2025. Google Expands ads in AI Overviews, AI Mode to desk- top. https://searchengineland.com/google-ads-ai-overviews-ai-mode-desktop- 455733. (May 2025). Search Engine Land. Accessed: 2026-01-19

    work page 2025

  2. [2]

    Amazeen and Bartosz W

    Michelle A. Amazeen and Bartosz W. Wojdynski. 2020. The Effects of Disclosure Format on Native Advertising Recognition and Audience Perceptions of Legacy and Online News Publishers.Journalism21, 12 (2020), 1965–1984. https://doi. org/10.1177/1464884918754829

    work page doi:10.1177/1464884918754829 2020

  3. [3]

    Janek Bevendorff, Matti Wiegmann, Martin Potthast, and Benno Stein. 2024. Product Spam on YouTube: A Case Study. In9th ACM SIGIR Conference on Human Information Interaction and Retrieval (CHIIR 2024), Min Zhang, Joemon Jose, and Laurianne Sitbon (Eds.). ACM, Sheffield, United Kingdon, 358–363. https://doi.org/10.1145/3627508.3638303

    work page doi:10.1145/3627508.3638303 2024

  4. [4]

    Leo Breiman. 2001. Random Forests.Machine Learning45, 1 (Oct. 2001), 5–32. https://doi.org/10.1023/A:1010933404324

    work page doi:10.1023/a:1010933404324 2001

  5. [5]

    Cadet, Priscilla G

    Fabienne T. Cadet, Priscilla G. Aaltonen, and Vahwere Kavota. 2017. The Ad- vertisement Value of Transformational & Informational Appeal on Company Facebook Pages.Marketing Management Journal27, 2 (jan 1 2017), 116–130. https://doi.org/10.63963/001c.151117

    work page doi:10.63963/001c.151117 2017

  6. [6]

    Fanny Fong Yee Chan. 2019. The Perceived Effectiveness of Overt Versus Covert Promotions.Journal of Product & Brand Management29, 3 (Aug. 2019), 321–334. https://doi.org/10.1108/JPBM-06-2018-1912

    work page doi:10.1108/jpbm-06-2018-1912 2019

  7. [7]

    Davit Davtyan and Isabella Cunningham. 2017. An Investigation of Brand Placement Effects on Brand Attitudes and Purchase Intentions: Brand Place- ments versus TV Commercials.Journal of Business Research70 (2017), 160–167. https://doi.org/10.1016/j.jbusres.2016.08.023

    work page doi:10.1016/j.jbusres.2016.08.023 2017

  8. [8]

    Avinava Dubey, Zhe Feng, Rahul Kidambi, Aranyak Mehta, and Di Wang. 2024. Auctions with LLM Summaries. InProceedings of the 30th ACM SIGKDD Confer- ence on Knowledge Discovery and Data Mining. ACM, Barcelona Spain, 713–722. https://doi.org/10.1145/3637528.3672022

    work page doi:10.1145/3637528.3672022 2024

  9. [9]

    Paul Dütting, Vahab Mirrokni, Renato Paes Leme, Haifeng Xu, and Song Zuo

    work page

  10. [10]

    InProceedings of the ACM Web Conference 2024 (WWW ’24)

    Mechanism Design for Large Language Models. InProceedings of the ACM Web Conference 2024 (WWW ’24). Association for Computing Machinery, New York, NY, USA, 144–155. https://doi.org/10.1145/3589334.3645511

    work page doi:10.1145/3589334.3645511 2024

  11. [11]

    Fabian Göbel, Anton Meyer, Balasubramani Ramaseshan, and Silke Bartsch

    work page

  12. [12]

    https://doi.org/10.1108/MIP-11- 2016-0212

    Consumer Responses to Covert Advertising in Social Media.Marketing Intelligence & Planning35, 5 (2017), 578–593. https://doi.org/10.1108/MIP-11- 2016-0212

    work page doi:10.1108/mip-11- 2017

  13. [13]

    Siyu Gong and Li Wang. 2023. Are Explicit or Implicit Appeals More Credible? The Congruence Effects of Green Advertising Appeals and Product Category on Consumers’ Evaluation.Current Psychology42, 33 (2023), 29035–29047. https: //doi.org/10.1007/s12144-022-03981-4

    work page doi:10.1007/s12144-022-03981-4 2023

  14. [14]

    Danny Goodwin. 2025. ChatGPT ads are coming - and they won’t look like Google Ads. https://searchengineland.com/chatgpt-ads-coming-some-point- 464388. (July 2025). Search Engine Land. Acessed: 2026-01-19

    work page 2025

  15. [15]

    Mohammad Taghi Hajiaghayi, Sébastien Lahaie, Keivan Rezaei, and Suho Shin

    work page

  16. [16]

    InProceedings of the 38th International Conference on Neural Information Processing Systems (NIPS ’24)

    Ad auctions for LLMs via retrieval augmented generation. InProceedings of the 38th International Conference on Neural Information Processing Systems (NIPS ’24). Curran Associates Inc., Red Hook, NY, USA, Article 585, 36 pages. https://doi.org/10.48550/arXiv.2406.09459

    work page doi:10.48550/arxiv.2406.09459

  17. [17]

    Johannes Kiesel, Çağrı Çöltekin, Marcel Gohsen, Sebastian Heineking, Maximilian Heinrich, Maik Fröbe, Tim Hagen, Mohammad Aliannejadi, Sharat Anand, Tomaž Erjavec, Matthias Hagen, Matyáš Kopp, Nikola Ljubešić, Katja Meden, Nailia Mirzakhmedova, Vaidas Morkevičius, Harrisen Scells, Moritz Wolter, Ines Zelch, Martin Potthast, and Benno Stein. 2025. Overview...

    work page 2025

  18. [18]

    Adam: A Method for Stochastic Optimization

    Diederik P. Kingma and Jimmy Ba. 2015. Adam: A Method for Stochastic Opti- mization. (2015). arXiv:1412.6980 http://arxiv.org/abs/1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  19. [19]

    Brice Valentin Kok-Shun and Johnny Chan. 2025. Leveraging ChatGPT for Sponsored Ad Detection and Keyword Extraction in YouTube Videos. (2025). https://doi.org/10.48550/arXiv.2502.15102 arXiv:2502.15102

    work page doi:10.48550/arxiv.2502.15102 2025

  20. [20]

    2018.Principles of Marketing(17th ed.)

    Philip Kotler. 2018.Principles of Marketing(17th ed.). Pearson, Harlow, England

    work page 2018

  21. [21]

    Florian Leiser, Sven Eckhardt, Valentin Leuthe, Merlin Knaeble, Alexander Mäd- che, Gerhard Schwabe, and Ali Sunyaev. 2024. HILL: A Hallucination Iden- tifier for Large Language Models. InProceedings of the CHI Conference on Human Factors in Computing Systems. ACM, Honolulu HI USA, 1–13. https: //doi.org/10.1145/3613904.3642428

    work page doi:10.1145/3613904.3642428 2024

  22. [22]

    Ilya Loshchilov and Frank Hutter. 2017. Decoupled Weight Decay Regularization. (2017). arXiv:1711.05101 http://arxiv.org/abs/1711.05101

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  23. [23]

    Roseanne Luth. 2017. 3 Ways to Use Recipes in Your Marketing. https://www. brandingmag.com/2017/08/19/3-ways-to-use-recipes-in-your-marketing/. (Aug. 2017). Brandingmag. Accessed: 2026-01-19

    work page 2017

  24. [24]

    Tomás Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient Estimation of Word Representations in Vector Space. (2013). http://arxiv.org/ abs/1301.3781

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  25. [25]

    Tomás Mikolov, Ilya Sutskever, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Distributed Representations of Words and Phrases and Their Compositionality. In Proceedings of the 27th International Conference on Neural Information Processing Systems - Volume 2 (NIPS’13). Curran Associates Inc., Red Hook, NY, USA, 3111–

    work page 2013

  26. [26]

    https://dl.acm.org/doi/10.5555/2999792.2999959

    work page doi:10.5555/2999792.2999959

  27. [27]

    Barbara Mueller. 1987. Reflections of Culture: An Analysis of Japanese and American Advertising Appeals.Journal of Advertising research27, 3 (1987), 51–59. https://eric.ed.gov/?id=ED271776

    work page 1987

  28. [28]

    Wael Nuweihed and Olivier Trendel. 2024. The role of informational versus transformational ad appeals in building consumer-based brand equity for low involvement products.Journal of Marketing Theory and Practice32, 4 (2024), 579–598. https://doi.org/10.1080/10696679.2023.2249213

    work page doi:10.1080/10696679.2023.2249213 2024

  29. [29]

    Shintaro Okazaki, Barbara Mueller, and Charles R. Taylor. 2010. Measuring Soft- Sell versus Hard-Sell Advertising Appeals.Journal of Advertising39, 2 (2010), 5–20. https://doi.org/10.2753/JOA0091-3367390201

    work page doi:10.2753/joa0091-3367390201 2010

  30. [30]

    Fabian Pedregosa, Gaël Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, et al. 2011. Scikit-learn: Machine Learning in Python.Journal of Machine Learning Research12, 85 (2011), 2825–2830. http://jmlr.org/papers/ v12/pedregosa11a.html

    work page 2011

  31. [31]

    Perplexity. 2024. Why we’re experimenting with advertising. https://www. perplexity.ai/hub/blog/why-we-re-experimenting-with-advertising. (Dec. 2024). Perplexity.ai. Accessed: 2026-01-19

    work page 2024

  32. [32]

    Louvins Pierre. 2024. The Effect of Covert Advertising Recognition on Consumer Attitudes: A Systematic Review.Journal of Marketing Communications30, 8 (2024), 1077–1098. https://doi.org/10.1080/13527266.2023.2184851

    work page doi:10.1080/13527266.2023.2184851 2024

  33. [33]

    Colin Porlezza. 2017. Digitaler Journalismus zwischen News und Native Adver- tising - Risiken und Nebenwirkungen einer heiklen Beziehung. InAbbruch - Umbruch - Aufbruch, Werner A. Meier (Ed.). Nomos Verlagsgesellschaft mbH & Co. KG, Baden-Baden, Germany, 249–270. https://doi.org/10.5771/9783845276663-249

    work page doi:10.5771/9783845276663-249 2017

  34. [34]

    L. A. Ramshaw and M. P. Marcus. 1999. Text Chunking Using Transformation- Based Learning. InNatural Language Processing Using Very Large Corpora, Susan Armstrong, Kenneth Church, Pierre Isabelle, Sandra Manzi, Evelyne Tzouker- mann, and David Yarowsky (Eds.). Springer Netherlands, Dordrecht, 157–176. https://doi.org/10.1007/978-94-017-2390-9_10

    work page doi:10.1007/978-94-017-2390-9_10 1999

  35. [35]

    Sravana Reddy, Yongze Yu, Aasish Pappu, Aswin Sivaraman, Rezvaneh Rezapour, and Rosie Jones. 2021. Detecting Extraneous Content in Podcasts. InProceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume. Association for Computational Linguistics, Online, 1166–1173. https://doi.org/10.18653/v1/20...

    work page doi:10.18653/v1/2021.eacl-main.99 2021

  36. [36]

    Radim Řehůřek and Petr Sojka. 2010. Software Framework for Topic Modelling with Large Corpora. InProceedings of the LREC 2010 Workshop on New Challenges for NLP Frameworks. European Language Resources Association (ELRA), Valletta, Malta, 45–50. https://doi.org/10.13140/2.1.2393.1847

    work page doi:10.13140/2.1.2393.1847 2010

  37. [37]

    Kya Sainsbury-Carter. 2025. Transforming the future of audience engagement. https://about.ads.microsoft.com/en/blog/post/march-2025/transforming-the- future-of-audience-engagement?s_cid=gl-ob-impc-gai-src_wpage-sub_oooc- cam_accelerate-flx_copilotlp. (May 2025). Microsoft Advertising. Accessed: 2026-01-19. Detecting RAG Advertisements Across Advertising S...

    work page 2025

  38. [38]

    Schauster, Patrick Ferrucci, and Marlene S

    Erin E. Schauster, Patrick Ferrucci, and Marlene S. Neill. 2016. Native Ad- vertising is the New Journalism: How Deception Affects Social Responsibility. American Behavioral Scientist60, 12 (2016), 1408–1424. https://doi.org/10.1177/ 0002764216660135

    work page 2016

  39. [39]

    Sebastian Schmidt, Ines Zelch, Janek Bevendorff, Benno Stein, Matthias Hagen, and Martin Potthast. 2024. Detecting Generated Native Ads in Conversational Search. InCompanion Proceedings of the ACM Web Conference 2024. ACM, Singa- pore Singapore, 722–725. https://doi.org/10.1145/3589335.3651489

    work page doi:10.1145/3589335.3651489 2024

  40. [40]

    Fidji Simo. 2026. Our approach to advertising and expanding access to Chat- GPT. https://openai.com/index/our-approach-to-advertising-and-expanding- access/. (Jan. 2026). OpenAI. Accessed: 2026-01-211

    work page 2026

  41. [41]

    Peter Snyder, Antoine Vastel, and Ben Livshits. 2020. Who Filters the Fil- ters: Understanding the Growth, Usefulness and Efficiency of Crowdsourced Ad Blocking. InAbstracts of the 2020 SIGMETRICS/Performance Joint Interna- tional Conference on Measurement and Modeling of Computer Systems, Edmund Yeh, Athina Markopoulou, and Y. C. Tay (Eds.). ACM, Boston,...

    work page doi:10.1145/3393691.3394228 2020

  42. [42]

    Jennifer Sor. 2025. Sam Altman touts ChatGPT’s 800 million weekly users, double all its main competitors combined. https://www.businessinsider.com/chatgpt- users-openai-sam-altman-devday-llm-artificial-intelligence-2025-10. (Oct. 2025). Business Insider. Accessed: 2026-01-19

    work page 2025

  43. [43]

    Curry, and Sven Seuken

    Ermis Soumalias, Michael J. Curry, and Sven Seuken. 2024. Truthful Aggregation of LLMs with an Application to Online Advertising. (2024). https://doi.org/10. 48550/ARXIV.2405.05905 arXiv:cs.GT/2405.05905

    work page arXiv 2024

  44. [44]

    Stone, Dexter C

    Philip J. Stone, Dexter C. Dunphy, and Marshall S. Smith. 1966.The General Inquirer: A Computer Approach to Content Analysis.M.I.T. Press, Oxford, England. Pages: 651

    work page 1966

  45. [45]

    Abu Sultan. 2025. OpenAI projects 220 million paying ChatGPT users by 2030, The Information Reports. https://www.reuters.com/technology/openai-projected- least-220-million-people-will-pay-chatgpt-by-2030-information-2025-11-26/. (Nov. 2025). Reuters. Accessed: 2026-01-19

    work page 2025

  46. [46]

    Demetris Vrontis, Anna Makrides, Michael Christofi, and Alkis Thrassou. 2021. Social media influencer marketing: A systematic review, integrative framework and future research agenda.International Journal of Consumer Studies45, 4 (July 2021), 617–644. https://doi.org/10.1111/ijcs.12647

    work page doi:10.1111/ijcs.12647 2021

  47. [47]

    Wojdynski

    Bartosz W. Wojdynski. 2016. The Deceptiveness of Sponsored News Articles: How Readers Recognize and Perceive Native Advertising.American Behavioral Scientist60, 12 (2016), 1475–1491. https://doi.org/10.1177/0002764216660140

    work page doi:10.1177/0002764216660140 2016

  48. [48]

    Wojdynski and Nathaniel J

    Bartosz W. Wojdynski and Nathaniel J. Evans. 2020. The Covert Advertising Recognition and Effects (CARE) Model: Processes of Persuasion in Native Ad- vertising and Other Masked Formats.International Journal of Advertising39, 1 (2020), 4–31. https://doi.org/10.1080/02650487.2019.1658438

    work page doi:10.1080/02650487.2019.1658438 2020

  49. [49]

    Ines Zelch, Matthias Hagen, and Martin Potthast. 2024. A User Study on the Acceptance of Native Advertising in Generative IR. InProceedings of the 2024 ACM SIGIR Conference on Human Information Interaction and Retrieval. ACM, Sheffield United Kingdom, 142–152. https://doi.org/10.1145/3627508.3638316

    work page doi:10.1145/3627508.3638316 2024