Quotation Addition

Quotation Addition¹ is the source-content modification method tested in Aggarwal et al. 2023's GEO paper that produced the largest measured Table 1 standalone PAWC value among the 9 methods tested²: the method actively rewrites content to include direct quotations from authorities relevant to the topic. In the paper's main GEO-bench evaluation, Quotation Addition scored PAWC 27.2 vs the no-modification baseline of 19.3 (the position-adjusted "Overall" column of Table 1); the mathematically derived relative gain is +41%, matching the paper's headline "up to 40%" and its "30-40% relative improvement" range for the verbatim named top-3 methods (Cite Sources, Quotation Addition, and Statistics Addition). Quotation Addition appears in that named top-3 with the highest standalone PAWC; Statistics Addition and Cite Sources complete the named group. Fluency Optimization is 3rd by standalone PAWC (24.7) but is not in the paper's named top-3.

The practitioner discipline framing extends the paper's one-shot intervention into a habitual writing technique: do not just add quotations during a citation-optimization pass; write the content with sourced direct quotations as a standard habit. This extension is a glossary inference from the intervention finding, not a paper conclusion. The paper applied the method as an LLM-prompted rewrite over source content; the editorial habit of writing with quotations from the first draft is the practitioner discipline interpretation of the same finding.

Status in 2026

Quotation Addition appears in many 2026 content marketing and GEO guides as a recommended technique, often without the paper's empirical context (specific named-guide endorsements vary and are not cataloged here). The paper's actual finding is narrower: under specific 2023 testbed conditions (GPT-3.5-turbo, top-5 Google sources, temperature=0.7, 5 responses per query), the LLM-prompted Quotation Addition intervention raised PAWC from 19.3 to 27.2. Aggarwal et al. describe the intervention as one that "modifies content to include authoritative quotations relevant to the topic", tested as a one-shot LLM rewrite rather than as a habitual writing pattern. Whether the 27.2 PAWC result for Quotation Addition reproduces on 2026 commercial AI engines (ChatGPT-5, Perplexity, Claude, Copilot, Gemini) has not been isolated by public study; the score is experimental signal under the paper's specific testbed, not a citation-rate multiplier for any current engine.

Quotation Addition increases PAWC under Aggarwal's experimental conditions; it does not guarantee citation by ChatGPT, Perplexity, Claude, Copilot, or Gemini. Authority of the original quotation source, freshness, source quality, and the topical context where the quotation appears all remain independent signals beyond the quotation-addition discipline alone.

Counter-evidence (C-SEO Bench 2025): A follow-up benchmark³ specifically tested Quotation Addition (referenced as "Quotes" in the paper) across question-answering and product-recommendation tasks on multiple domains under varying multi-actor adoption rates. See C-SEO Bench for the full multi-actor benchmark methodology, the comparison to traditional retrieval-ranking SEO, and the zero-sum framing. The C-SEO Bench result does not invalidate Aggarwal's PAWC 27.2 measurement for Quotation Addition (that remains valid for the 2023 single-actor synthetic testbed), but it sets an empirical upper bound: Quotation Addition in production-realistic multi-domain multi-actor conditions appears to show minimal or negative effect, not the ~41% lift seen on the 2023 benchmark.

How to apply

The Aggarwal paper supports active rewriting to add direct quotations from real authorities relevant to the topic. It does not support quote-padding or decorative quotations. The practical writing rules:

• Find authority quotations that genuinely support the claim you are making: scan your draft for any sentence that paraphrases an established voice (research findings, vendor documentation, recognized practitioners) and consider replacing the paraphrase with a verbatim quote from the source. Example: "researchers showed that dense retrieval can replace BM25 for open-domain QA" can become "Karpukhin et al. (2020) report that their 'dense retriever outperforms a strong Lucene-BM25 system greatly by 9%-19% absolute in terms of top-20 passage retrieval accuracy' (EMNLP 2020, arXiv:2004.04906)."⁴

• Verify the quotation against the original source: copy the wording exactly, not from memory. Confirm the source URL and publication date. A misquoted or out-of-context quotation undermines editorial trust with readers, and may be less useful to retrieval or citation systems that compare quoted text against the cited source.

• Attribute fully: author name + role or affiliation + publication or paper + date or year. Bare quotation marks without attribution do not activate the citation-friendly pattern; the attribution itself is part of what makes the quote useful for an AI engine to ground from.

• Lead with the substantive content, not setup: a paragraph that opens with several sentences of context before reaching the quotation buries the cite-worthy content. Either journalistic style (quotation first, then attribution) or scholarly style (attribution first, then quotation) works; concept density at the paragraph opening matters more than the specific ordering. The primary motivation is human readability and chunk-boundary robustness; if a passage is summarized or extracted, the substantive content is what gets surfaced. Liu et al. 2023 "Lost in the Middle" (arXiv:2307.03172) observed primacy and recency attention patterns in transformer processing, but whether this translates to citation selection by 2026 commercial AI engines has not been vendor-documented. What to skip:

• Padding pages with decorative quotations that do not support a specific claim.

• Self-quotation or quoting in-house material as "authority". The paper tested external authority sources. (Internal /terms/ cross-links are navigation aids, not authority citations; the external authority signal comes from Sources and academic footnotes.)

• Fabricated or out-of-context quotations attributed to real authorities. AI engines retrieving from the same corpus may cross-check quotations against the cited source; misrepresenting the original risks contradiction and downweighting. Treat as editorial caution, not a paper finding.

How it relates to other concepts

• The underlying intervention is one of nine GEO methods tested in Aggarwal et al. 2023. The paper's top 4 (Quotation Addition, Statistics Addition, Fluency Optimization, Cite Sources) cover the strongest content-level interventions isolated under its test conditions. The strongest pairwise combination the paper isolated was Fluency Optimization plus Statistics Addition (§5.3 Figure 4 heatmap, 200-example subset), outperforming any single method by more than 5.5%.
• Closely related to Cite Sources Optimization: both methods overlap because most authority quotations carry source citation. The paper measured them separately.
• Often co-applied with Statistical Density (the entry covering Aggarwal's Statistics Addition method): a quotation that backs up a numerical claim activates both treatments simultaneously. Example: citing Ahrefs' March 2026 study finding that "37.9% of Google AI Overview cited pages also ranked in the organic top 10" (based on 4M URLs across 863K SERPs)⁵ is both a sourced quantitative claim and a verified-link citation.
• Operationally similar to passage-level optimization: quotations make passages self-contained (the passage carries its own external evidence) and cite-ready (a clean cite-able chunk for AI engines that extract passage-level sources).
• May contribute to broader cite-ability; any effect on citation velocity should be measured over time per page rather than assumed from content discipline alone.