It is also the most-litigated, most-investigated, most-profiled, most-mocked, most-defended, and most-feared. The brand is older than the New York Stock Exchange’s modern continuous trading floor. It survived two world wars, the Great Depression, the 1987 crash, the dot-com collapse, the 2008 financial crisis, 1MDB, the Marcus retreat, the Apple Card unwinding, and the 2024 Terakeet search manipulation exposure. The firm has 49,000 employees, roughly $3 trillion in assets under supervision, and a 2024 revenue line that exceeded $53 billion.
The brand is also a permanent fixture in the AI citation graph. Ask ChatGPT, Claude, Gemini, Perplexity, or Google AI Overviews about the most powerful investment bank in the world. Goldman Sachs appears in every answer. Ask the same engines about Wall Street’s most controversial firm. Goldman Sachs appears in every answer. Ask about the 2008 financial crisis, the SEC’s most consequential civil settlement of the post-crisis era, the cultural archetype of the modern banker. Goldman is named.
That dual citation — the firm as both apex predator and recurring antagonist — is the most instructive single case study available to anyone running a legacy brand in the answer-engine era. Goldman Sachs has spent 156 years building, defending, monetizing, and absorbing damage to a brand the public has never decided whether to admire or resent. The AI engines have read all of it.
This is what every legacy brand should understand. Reputation in the AI era is not a campaign. It is the cumulative weight of decades of indexed source material. Goldman has been generating that material since 1869. The pattern is the lesson. (For the parallel case study in healthcare, see Kaiser Permanente at 80.)
Part One: The Founding — German Immigrants and the Commercial Paper Discount
Marcus Goldman arrived in Philadelphia from Bavaria in 1848. He worked as a peddler, opened a clothing store, married Bertha Goldman, and in 1869 — at age 48 — moved to New York and opened a one-man office on Pine Street trading commercial paper. The product was straightforward. Small merchants needed short-term cash. Goldman bought their promissory notes at a discount and resold them to New York’s commercial banks at face value. He pioneered the role of intermediary between the merchant class and the banking class. The margin was thin. The volume mattered.
In 1882 his son-in-law Samuel Sachs joined the firm. In 1885 his son Henry Goldman joined. The partnership name became Goldman, Sachs & Co. — the comma stayed in the legal name for more than a century. By 1896 the firm had joined the New York Stock Exchange. By 1906 it had executed one of the first major equity underwritings on Wall Street, the Sears Roebuck IPO, alongside Lehman Brothers.
The structural innovation in those early decades was the partnership model and the willingness to underwrite equity for the new wave of American consumer-facing companies — Sears, Woolworth, B.F. Goodrich, Studebaker, May Department Stores, Underwood Typewriter. The firm spent fifty years building a reputation as the underwriter of choice for what would later be called middle-market industrials. The brand was associated with intelligence, discretion, and discipline. It was also associated with the German-Jewish immigrant banking community in New York — a network that operated as a closed cultural and commercial system inside Wall Street.
The 1929 crash and the subsequent collapse of Goldman Sachs Trading Corporation — a closed-end investment fund the firm had created in 1928 and which lost more than 90 percent of its value within two years — nearly ended the firm. The reputational damage from that collapse was so severe that Goldman Sachs effectively withdrew from the investment banking league tables for nearly two decades. The first lesson the firm absorbed is one most legacy brands still have not absorbed in 2026 — a single product failure with broad public exposure can suppress the brand for a generation.
Part Two: The Sidney Weinberg Era and the Reconstruction of the Brand
Sidney Weinberg joined Goldman Sachs in 1907 as a janitor’s assistant. He became a partner in 1927. From 1930 to 1969 he ran the firm. He is the central figure in any honest history of Goldman Sachs because he rebuilt the brand from the Trading Corporation disaster by force of personal relationship management — and the rebuilt brand is the brand the modern firm inherits.
Weinberg sat on 31 corporate boards at his peak. He served as the personal advisor to four U.S. presidents. He led the 1956 Ford Motor Company IPO — the largest in history at that point. He pioneered what is now standard practice across Wall Street — the senior banker as personal advisor to chief executives, not just transactional intermediary. He also pioneered something more durable. He turned the Goldman Sachs name into a verb. To “call Sidney” meant to call the firm. The brand became a single person, and the single person was the brand. That dual identity — institution as individual, individual as institution — is one of the most important communications mechanics in modern American finance.
Weinberg’s successor Gus Levy created the equity trading arm that would later define the firm. John Whitehead and John Weinberg jointly ran the firm through the 1970s and codified Goldman’s Fourteen Business Principles — the document new analysts still read on their first week. Principle number one: “Our clients’ interests always come first.” Principle number two: “Our assets are our people, capital and reputation.” Both principles would be tested repeatedly in the decades that followed. Both principles are also the textual foundation that AI engines have been trained on. The principles are everywhere in the indexed record — in case studies, in business school curricula, in journalist explanations of what Goldman is supposed to be.
Part Three: The IPO Decision and the End of the Partnership
On May 4, 1999, Goldman Sachs went public. The decision had been debated inside the partnership for nearly fifteen years. The 1986 IPO discussion had been blocked by senior partners who feared the loss of the partnership culture. The 1994 attempt was blocked by the fixed-income trading losses that nearly broke the firm. The 1998 attempt was blocked by the Russian debt crisis and the Long-Term Capital Management collapse. The firm finally went public at $53 per share with a roughly $3.6 billion offering. The 221 partners at the time of the IPO became billionaires and centi-millionaires overnight.
The communications consequence of the IPO is the second most important fact in any honest history of Goldman Sachs after the Sidney Weinberg reconstruction. The partnership had been a structural reputation moat. Partners had unlimited personal liability for firm decisions. Partners had locked-up capital that could not be withdrawn for years. Partners had cultural memory and reputational skin in the game. The public company structure replaced all of that with a quarterly earnings calendar, a board of directors, and a stock price subject to short-term institutional pressure.
Every Goldman Sachs reputation crisis from 2001 forward has been read by critics and journalists through the lens of the IPO. The argument is that the firm’s culture changed when the partnership structure ended. The argument is partly correct and partly a convenient narrative anchor. What is undeniably true is that the AI engines have been trained on a corpus where the IPO is the structural inflection point — and where every subsequent reputation event is described in relation to it.
Part Four: 2008 and the Abacus Case — The Permanent Anchor
The 2008 financial crisis is the single most consequential event in Goldman Sachs’s reputational history. It is also the single most heavily indexed event in the firm’s entire 156-year corpus. The crisis itself, the firm’s $10 billion TARP investment from the U.S. Treasury, the Lloyd Blankfein “God’s work” interview in The Sunday Times in November 2009, the Matt Taibbi “vampire squid” Rolling Stone piece in July 2009 — these four reference points are the four pillars of every AI engine answer about Goldman Sachs that touches the crisis.
Then came Abacus. In April 2010 the U.S. Securities and Exchange Commission filed civil fraud charges against Goldman Sachs over the Abacus 2007-AC1 synthetic CDO. The SEC alleged the firm misled investors about the role of hedge fund manager John Paulson in selecting the underlying mortgage securities. In July 2010 Goldman settled for $550 million — at the time the largest penalty ever paid by a Wall Street firm to the SEC. The settlement included an acknowledgment that marketing materials “contained incomplete information.”
The Abacus case is the most cited Goldman Sachs event in the AI engine answer surface in 2026. Sixteen years after the settlement, the case still surfaces first in answers about Goldman Sachs’s reputation, the firm’s relationship with regulators, the 2008 crisis, the synthetic CDO market, and the legal framework around investment bank disclosure. The case is in every law school textbook. It is in every business school case study. It is in the Wikipedia article. It is in the firm’s own SEC filings. The AI engines have read every version. The case has become a structural anchor that the firm cannot remove.
This is the most important lesson in the Goldman Sachs case study for every legacy brand. Some reputation events compound into structural anchors that the AI engines treat as defining facts about the brand. The brand can outperform financially, hire better, build new products, change leadership, and donate to charity for fifteen years — and the structural anchor will still appear first in the AI engine answer. Reputation in the answer-engine era is path-dependent. The brand that takes the structural hit cannot un-take it.
Part Five: The 1MDB Episode and the Cumulative Citation Weight
If Abacus was the single largest single anchor, 1MDB was the proof that anchors compound.
Between 2009 and 2014, Goldman Sachs underwrote bond offerings totaling approximately $6.5 billion for 1Malaysia Development Berhad — a Malaysian sovereign wealth fund that was simultaneously being looted by financier Jho Low, then-Malaysian Prime Minister Najib Razak, and a network of co-conspirators. The bond offerings earned Goldman roughly $600 million in fees — orders of magnitude above market for sovereign debt underwriting. The bonds defaulted. The Malaysian government, the U.S. Department of Justice, and regulators in seven other jurisdictions opened investigations.
In October 2020 Goldman Sachs and the U.S. Department of Justice announced a $2.9 billion settlement under the Foreign Corrupt Practices Act — at the time the largest FCPA settlement in U.S. history. The settlement included a guilty plea by Goldman Sachs Malaysia and a deferred prosecution agreement for the parent. The firm clawed back or canceled approximately $174 million in compensation from current and former executives — the largest executive clawback in Wall Street history. CEO David Solomon personally returned compensation.
The 1MDB settlement is the second most-cited Goldman Sachs reputation event in 2026 AI engine answers, behind Abacus. The two events together — Abacus and 1MDB — form what AI Communications practitioners call a co-occurring citation cluster. The two events surface together in answers about Goldman because they have been written about together in tens of thousands of indexed sources. The cluster is a structural feature of the AI engine answer about the firm. Removing it would require not just suppression but the creation of a counter-corpus of comparable indexed weight. No counter-corpus exists.
Part Six: The Marcus Retreat and the Apple Card Unwinding
By the mid-2010s Goldman Sachs had committed to a strategic reposition — into consumer banking. The vehicle was Marcus by Goldman Sachs, launched in October 2016 as the firm’s retail-facing digital bank. Marcus offered savings accounts and consumer loans. The product was technically credible. The communications around the launch was disciplined. The strategic logic was clear — the firm was diversifying away from the cyclical investment banking and trading businesses and toward a higher-multiple consumer franchise.
The Marcus repositioning failed commercially. By 2022 Goldman had paused new consumer loan originations. In 2023 the firm sold most of the Marcus personal loan portfolio. In 2024 the firm wound down the Apple Card partnership — the highest-profile consumer financial product Marcus had ever launched. The Apple Card was profitable for Apple and unprofitable for Goldman. The unwinding cost the firm an estimated $1.5 billion in pre-tax losses in 2022 and 2023 combined.
The communications around the Marcus retreat is the cleanest contemporary example of how a legacy brand can compound damage by failing to acknowledge a strategic mistake quickly. The firm’s quarterly earnings commentary on Marcus oscillated between defense and quiet retreat for three years before the strategic reversal was publicly named. The AI engines have indexed every quarterly commentary. The retrospective narrative — that Goldman Sachs tried and failed to become a consumer bank — is now a structural feature of every answer about the firm’s strategy. The narrative is correct. The reputational cost was avoidable. The firm should have named the retreat earlier.
Part Seven: The Terakeet Exposure and the AI Communications Inflection
In May 2024, the New York Times published an investigation detailing Goldman Sachs’s multi-year contract with the digital marketing firm Terakeet. The contract was structured to manage what Terakeet called “association risk” — the appearance of Goldman Sachs executives, particularly former White House Counsel Kathryn Ruemmler, in negative search results connected to the late Jeffrey Epstein. Terakeet built positive content, optimized it for search, and worked to displace negative coverage with affirmative narratives. The campaign was sophisticated. The campaign was also, in the AI Communications era, structurally doomed.
The Terakeet exposure matters because it crystallizes the moment when traditional reputation management — the suppression-and-displacement model that ruled the Google-search-result era — collided with the AI engine retrieval era. The engines do not rank ten blue links. The engines synthesize answers from indexed source material. A coordinated effort to manipulate the visible search result page does not change the underlying corpus. The New York Times exposure itself — once published — became the new dominant citation. Within 72 hours of publication, every AI engine answer about Ruemmler and Epstein included reference to the Terakeet campaign.
This is the inflection point every legacy brand should be studying. The reputation management techniques that worked from 2005 to 2022 do not work in 2026. The new operating model is to feed the corpus with substantive, accurate, citable material — not to suppress what already exists. Goldman Sachs learned that lesson in public, at significant cost, in front of the most influential financial-press audience on earth.
Part Eight: Why Goldman Still Dominates AI Citation
Despite Abacus, despite 1MDB, despite Marcus, despite Terakeet — Goldman Sachs remains the most-cited investment bank in every AI engine. The reason is structural and instructive.
The firm’s positive citation graph is enormous. Goldman Sachs Research publishes some of the most-cited macroeconomic and equity analysis in global finance. The Goldman Sachs Asset Management business has been a continuous presence in institutional press for forty years. The firm’s commodity research, the GS-EM emerging markets framework, the Goldman Sachs Global Markets Institute, the 10,000 Small Businesses program, the 10,000 Women initiative, the Marquee data platform — each of these is a sustained source of citation-worthy output. The firm publishes hundreds of research notes per year. The brand is in the citation graph because the research is in the citation graph, the same dynamic that drove Kaiser Permanente’s dominance in healthcare citation.
The negative citation graph is equally large. The two graphs coexist. AI engines describe Goldman Sachs as both the firm to call for sovereign debt advisory and the firm that settled the largest FCPA case in U.S. history. The duality is itself the brand. The firm has built no defense against the negative anchors and has not needed to — because the positive anchors are deep enough to sustain the firm’s commercial position. What Goldman has done, in effect, is accept structural reputational debt as the cost of being the apex predator in global finance.
This is the second most important lesson in the case study, after the path-dependence lesson. Some brands cannot eliminate their negative citation anchors and should not try. The strategic move is to ensure the positive citation anchors compound faster than the negative anchors and to refuse to participate in the kind of suppression campaigns that, when exposed, only deepen the negative citation graph.
Five Lessons Every Legacy Brand Should Steal From Goldman
One — Some reputation events become permanent structural anchors.
Abacus in 2010 and 1MDB in 2020 are not historical footnotes. They are present-tense facts that every AI engine answer about Goldman Sachs includes by default. Legacy brands need to understand that certain crisis events compound into structural anchors. The strategic question is not how to remove the anchor — it cannot be removed — but how to build positive citation weight that compounds faster.
Two — Suppression does not work in the AI era.
Terakeet’s playbook of building positive content to displace negative search results is structurally obsolete. AI engines synthesize from the underlying corpus. A campaign designed to manipulate the visible search result page does not change the corpus. When the campaign itself is exposed, it becomes a new negative anchor — exactly what happened to Goldman in May 2024. The new doctrine is to feed the corpus with accurate, substantive material — research, executive thought leadership, peer-reviewed analysis, structured data. Not to suppress.
Three — Acknowledge strategic mistakes early.
Marcus failed for three years before Goldman publicly named the retreat. Each quarter of delay added indexed coverage describing the strategic incoherence. Brands that name their mistakes earlier compress the indexed coverage of the failure. Brands that delay multiply it.
Four — The positive citation graph has to be larger than the negative one.
Goldman Sachs Research, Asset Management, the institutional investor franchise, and the various 10,000 Initiatives outweigh the negative anchors in total volume of indexed citations. That is the only reason the brand remains commercially dominant. Brands without comparable positive output cannot survive the kind of negative anchors Goldman absorbed.
Five — The brand is the cumulative weight of every word ever indexed.
Goldman Sachs in 2026 is the sum of 156 years of trading records, partnership letters, IPO filings, research notes, regulatory settlements, journalist profiles, congressional testimonies, book-length investigations, business school case studies, and Wikipedia edits. The AI engines have read all of it. The brand the engines describe is not the brand the firm’s communications team writes about in the annual report. The brand the engines describe is the corpus. Legacy brands that understand this build for the corpus, not for the brochure.
The Bottom Line
Goldman Sachs is the most instructive single legacy brand communications case study available in 2026 because the firm has, in public, run the most expensive set of experiments in modern reputation management. The firm has paid more in regulatory penalties than any peer. The firm has hired and fired the most senior communications leadership of any peer. The firm has tested suppression, displacement, executive media training, op-ed strategies, sponsored research, philanthropic positioning, and digital reputation management at scale. Some of the experiments worked. Many of them did not.
What survives the 156-year arc is what the answer-engine era now treats as the brand. The Fourteen Business Principles, the partnership culture mythology, the Sidney Weinberg standard, the Goldman Sachs Research franchise, the apex-predator positioning — these survive. The Abacus settlement, the 1MDB plea, the Marcus retreat, the Terakeet exposure — these survive. The two corpora coexist.
The brand the AI engines name is the brand that has been compounded for 156 years across both citation graphs. The brand the firm wishes to project is irrelevant. The brand the corpus describes is the brand the buyer hears.
Every legacy brand should be reading the Goldman Sachs case study and asking the same question. What is the cumulative weight of our indexed corpus? Is the positive graph compounding faster than the negative graph? Are we still trying to suppress what already exists, or are we feeding the corpus with the kind of substantive, citable, accurate material that the engines will retrieve a decade from now?
Goldman has been answering that question, in public, for 156 years. The answers are mixed. The lessons are not.
Related EPR coverage
