Discover Appears in 46.7% of AI Answers. Here's Our Five-Step Plan to Improve Its AI Visibility.

Mark HuntleyBy Mark HuntleyFounder and CEO
11 minutes read

Key Takeaways

  • Discover appears in 46.75% of benchmark AI answers and is recommended in 29.43%, but first-place placements are rare.
  • The main issue is not reach; it is turning mentions into stronger recommendations and top-ranked placements.
  • Most disputed claims center on Discover Home Loans, including ownership and whether the business is still operating.
  • The plan prioritizes primary-source verification, then page briefs, publisher checks, and buyer-question content based on observed answer patterns.

Answer Capsule

In November 2026, Discover appeared in 467 of 999 AI answers included in the Best Banks benchmark (46.75%). It received 294 recommendations (29.43%), with 147 Top 3 placements and 12 first-place recommendations. A mention means the company appeared in an answer; a recommendation means the answer affirmatively endorsed it.

Our plan compares the company with its competitors, checks conflicting AI claims against reliable sources, and identifies website pages, third-party sources and buyer questions to address. This edition establishes a research baseline. It does not claim a change from earlier editions measured on a different basis. Conflicting AI claims require verification before corrections are proposed.

Current AI visibility

Questions This Section Answers

  • How do Discover mentions, recommendations and placements compare with its Best Banks peers?

Discover recorded 467 mentions (46.75%) and 294 recommendations (29.43%) across 999 answers included in the benchmark. Top 3 placements numbered 147; first-place recommendations numbered 12. Framing among its mentions was 306 positive, 152 neutral and 9 negative.

Rank

Company

Mentions

Mention rate

Recommendations

Recommendation rate

Top 3

Top 3 rate

#1

#1 rate

1

Ally Bank

831

83.18%

606

60.66%

527

52.75%

204

20.42%

2

Capital One

798

79.88%

502

50.25%

424

42.44%

182

18.22%

3

Marcus by Goldman Sachs

513

51.35%

370

37.04%

251

25.13%

79

7.91%

4

Discover

467

46.75%

294

29.43%

147

14.71%

12

1.20%

5

Chase

575

57.56%

150

15.02%

87

8.71%

29

2.90%

6

Bank of America Corp.

448

44.84%

78

7.81%

27

2.70%

7

0.70%

7

Wells Fargo & Co.

362

36.24%

36

3.60%

6

0.60%

2

0.20%

8

U.S. Bancorp

260

26.03%

18

1.80%

4

0.40%

0

0.00%

Discover appears in roughly half of the benchmark answers and is recommended in a meaningful share of them, which puts it in the leading group of banks tracked here. The gap between being mentioned and being recommended is the practical issue: the brand is familiar to these AI systems but is not consistently endorsed.

Placement is the sharper constraint. Discover's appearances near the top of an answer are far fewer than its recommendations, and outright first-place finishes are rare. Reach is not the problem; being the answer the system leads with is. Proposed work should treat recommendation and position as separate goals rather than one visibility score.

Framing is largely positive or neutral, with only a small negative share. That suggests the opportunity is less about repairing reputation than about giving AI systems clearer, verifiable reasons to endorse Discover and to place it first.

What changed since last month

Questions This Section Answers

  • How did Discover counts and coverage differ across the available monthly snapshots?

Metric

Nov 2026

Mentions

467 / 999 (46.75%)

Recommendations

294 / 999 (29.43%)

Top 3 recommendations

147 / 999 (14.71%)

#1 recommendations

12 / 999 (1.20%)

Only months with verified results are shown. Monthly sample coverage is listed in Methodology.

This edition establishes a broader research baseline. Earlier editions used a narrower set of questions, so their figures are not treated as a performance trend.

This edition is the first broad baseline for Discover, built on the full set of screened questions rather than the narrower earlier set. The monthly figures should be read as a starting reference point, not as evidence of improvement or decline, because the question scope changed.

That means the observed totals describe where Discover stands in this wider sample. They do not show what moved between editions, and no cause should be attached to any difference from earlier figures. The archive remains useful for context only.

Future editions using the same screening scope can be compared against this baseline. Until then, the agency would treat this month as the anchor for scorecards and for tracking whether proposed work coincides with any change.

AI recommendation trend

Observed recommendation rate (%), Nov 2026

  • Discoverno change
    Nov 202629.43%

Separate selected monthly samples. Missing points are unavailable, not zero. Provider/prompt equivalence is unverified; rate movement does not demonstrate causation.

/ Take the next step

Want to Understand Your AI Citation Footprint?

We start every engagement with a full audit of how AI systems reference your brand today.

Measurable, Repeatable Programme

Build a durable foundation of credible citations that compounds over time and continues to influence AI answers as new queries emerge

Citation Architecture Review

Identify which high-authority community sources are and aren't working in your favour across AI platforms.

AI Visibility Audit

Understand exactly how LLMs are referencing your brand today and which sources are shaping those answers.

Step 1. Measure

Questions This Section Answers

  • What should the agency measure and verify first for Discover?

AI answers contained 4 conflicting claims about Discover: 3 critical, 1 high, 0 medium, 0 low. Of these, 4 were rated critical or high priority; the critical claims are included in that total. These ratings help prioritize fact-checking. They do not establish that a company or publisher made an error.

The fact-check list is short but weighted toward serious conflicts. All recorded disputed claims about Discover were rated critical or high priority, and most concern whether the home-loan business is operating and who owns it. Those are the questions a prospective customer would ask before acting.

A visibility comparison alongside this list would show the client where Discover stands relative to peers on the same questions, so effort goes to gaps that matter commercially rather than to every mention. The ratings prioritize checking; they do not establish that Discover or any publisher got something wrong.

Proposed work: build a baseline scorecard and a scoped fact-review register, with each disputed claim assigned an owner, an evidence requirement and a review status before any correction is considered.

Step 2. Make the website consistent

Questions This Section Answers

  • Which recorded claim differences about Discover should be verified before proposed page edits?

Review topic

Severity

Recorded AI claim A

Recorded AI claim B

who took over Discover Home Loans

high

Capital One acquired Discover Financial Services (which owned Discover Home Loans) in May 2025

Guaranteed Rate purchased the remains of Discover Home Loans after Discover exited mortgage origination in 2015

is it still in business / accepting applications

critical

partially available; some products stopped accepting new applications but still a viable option for home equity/refi

closed; no longer open for new applications, division shut down

is it still in business / accepting applications

critical

shut down; no longer accepting new applications

actively operating and available

is it still in business / accepting applications

critical

permanently closed, no longer in business

still a solid option, applications paused/limited

Both columns quote AI answers, not verified company facts. Resolve the entity, date and source before deciding whether a correction is warranted; different dates can explain an apparent conflict.

The disputed topics cluster around the home-loan business: whether it still accepts applications, whether it is closed, and which company took it over. The recorded AI answers contradict each other directly, and some conflicts may simply reflect different dates rather than error.

Resolving the entity and the timeline comes first. A page brief should be prepared only after current primary-source verification confirms the facts, and it should state plainly what the business does now and what changed. The brief should not adopt either AI claim as true in advance.

Proposed work: verify the ownership and operating status against primary sources, then draft conditional page briefs that answer the buyer's real question without overstating certainty.

Step 3. Align the external narrative

Questions This Section Answers

  • Which cited third-party pages are relevant to fact review for Discover?

Cited page selected for review

Citation appearances in all captured answers

Answers mentioning the company

Conflicting claims linked to this page

https://nerdwallet.com/mortgages/learn/discover-home-equity-loan

10

2

4

https://businessinsider.com/personal-finance/mortgages/discover-home-loans-review

8

1

3

https://themortgagereports.com/105346/discover-home-loans-review

6

1

3

https://fool.com/money/mortgages/discover-mortgage-review/

5

1

3

https://consumeraffairs.com/finance/discover-home-loans.html

5

0

3

https://thetruthaboutmortgage.com/discover-home-loans-review/

5

0

3

These pages were selected because they relate to the company in the captured answers. Their current contents still need checking before any correction request. A page linked to a conflicting claim is not necessarily its cause, and the same answer may contribute to more than one row.

Several third-party review pages are cited repeatedly in answers that also carry conflicting claims about Discover's home-loan business. That makes them sensible candidates for a closer look, but citation alone does not show that a page shaped an answer or caused a conflict.

Before any outreach, the agency would check each page's current contents and confirm which claims it actually makes. Requests to publishers should be evidence-backed and framed as corrections only where a verified inaccuracy exists; many pages may need no change at all.

Proposed work: a prioritized source-review list, a documented check of each page, and a clear record separating confirmed edits, no-change decisions, requests sent and responses received.

Step 4. Optimize priority company pages

Questions This Section Answers

  • Which observed Discover website pages should receive review briefs?

Company website page

Citation appearances in all captured answers

Benchmark answers citing the page

Recommending answers citing the page

https://discover.com/home-loans/

19

2

0

https://discover.com/credit-cards/card-smarts/credit-cards-for-beginners/

12

0

0

https://discover.com/credit-cards/card-smarts/first-credit-card/

10

0

0

https://discover.com/credit-cards/instant-use-credit-cards/

7

2

2

https://discover.com/credit-cards/student-credit-card/

3

0

0

https://discover.com/online-banking/

3

3

2

https://investorrelations.discover.com/newsroom/press-releases/press-release-details/2015/Discover-To-Close-Home-Loans-Business/default.aspx

3

0

0

https://discover.com/credit-cards/credit-cards-for-no-credit/

2

0

0

The company's website was cited 102 times across 48 distinct page addresses; the table shows 8 pages selected for review. Zeros in the answer columns mean no matching answers were recorded. A cited page is not automatically a page that accompanied a recommendation.

Discover's own pages are cited widely across captured answers, but the pattern is uneven. The home-loans page draws the most citation appearances, yet it is rarely cited in benchmark answers and never in recommending answers, which fits the disputed status of that business.

Credit-card education pages attract citations without accompanying recommendations, while a few product and banking pages appear in answers that do recommend Discover. The page briefs should therefore differ by page rather than follow one template.

Proposed work: page briefs built on approved facts and the audience questions observed in the research, with existing coverage checked first. A cited page is not automatically a page that drove a recommendation, so briefs should target clarity and verifiability rather than chase citation counts.

Step 5. Publish around buyer questions

Questions This Section Answers

  • Which observed buyer questions support content briefs for Discover?

Question observed in the research

Benchmark answers to this question

Company mentions

Company recommendations

What bank has no monthly fees?

3

3

3

What banks do not charge a monthly fee?

3

3

3

Who is the best bank to be with?

6

3

3

What banks are just online?

3

3

2

Is Marcus by Goldman Sachs a good bank?

4

2

2

Is Marcus by Goldman a good high-yield savings account?

3

2

2

What are the top 10 banking apps?

2

2

2

What bank can I open a checking account online with no deposit?

2

2

2

These are recorded answer observations, not search volume, verified demand, or proof of missing website content.

The observed questions are practical and comparison-driven: which banks charge no monthly fee, which are online-only, which banking apps rank highest, and how to open a checking account without a deposit. Discover already appears in answers to several of these, sometimes with a recommendation.

That makes these questions worth answering directly and clearly on the site, in language a buyer would use. The observations are recorded answer content, not search volume or proof that a page is missing, so existing coverage should be checked before proposing anything new.

Proposed work: buyer-question content briefs tied to the observed questions, each mapped to an existing page where one exists and flagged for review where coverage is thin.

Our 90-day action plan

Questions This Section Answers

  • How should the agency sequence the proposed Discover work and judge completion?

Period

Deliverable

Completion criterion

Days 1–30

Baseline scorecard, scoped fact-review register and page/source shortlist

Each record has an owner, evidence requirement and review status

Days 31–60

Approved page briefs and justified publisher requests

Separate confirmed edits, no-change decisions, requests and responses

Days 61–90

Buyer-content briefs and a fresh scorecard

Record work completed separately from measured answer outcomes and comparability limits

The first month establishes the baseline: a scorecard, a scoped fact-review register and a shortlist of pages and sources to examine. Each record needs an owner, an evidence requirement and a review status, so later decisions rest on verified facts rather than on AI claims.

The second month turns verified findings into approved page briefs and justified publisher requests, with confirmed edits, no-change decisions, requests and responses kept separate. The third month produces buyer-content briefs and a fresh scorecard.

Completed work will be reported separately from measured answer outcomes. The plan sets up the conditions for improvement; it does not promise a visibility gain, and any later comparison will carry the same scope limits noted in this report.

Methodology

Questions This Section Answers

  • How are Discover answer metrics and citation populations defined?

This report covers 2026-11. The exact dates when the original AI answers were collected have not been verified. The visibility rates use 999 answers that met the benchmark's inclusion rules. A company is recommended only when the answer both endorses it and uses positive wording. Top 3 and first-place counts are part of those recommendations, not extra recommendations.

An answer can mention or recommend several companies, so company counts cannot be added to find the number of distinct answers. The net sentiment score is positive minus negative mentions, divided by all company mentions; it describes answer wording, not customer satisfaction.

The wider citation analysis covers 1334 captured answers and 8528 appearances of cited URLs. Citation appearances across all captured answers differ from the number of benchmark answers citing a page and the number that also recommend a company. The current contents of cited pages and disputed company claims still need independent verification.

Monthly sample coverage:

Reporting month

Answers included

AI services covered

Services

Nov 2026

999

6

ChatGPT, Copilot, Gemini, Google AI Mode, Google AI Overviews, Perplexity

July 2026 measured a different set of companies, so it is excluded from this comparison. August 2026 measured a different set of companies, so it is excluded from this comparison. September 2026 measured a different set of companies, so it is excluded from this comparison. Earlier editions (October 2026) used a narrower set of questions. This edition includes all captured answers screened as relevant and establishes a broader baseline. Earlier figures remain in the archive; they are not used to claim a change in performance.

Future editions using the same screening scope can be compared against this baseline.

The benchmark covers a fixed set of answers that met inclusion rules, and the collection dates for the original AI answers have not been verified. A company counts as recommended only when an answer endorses it with positive wording; top placements sit inside those recommendations rather than adding to them.

Answers can mention several companies, so counts cannot be summed into distinct answers. Sentiment describes answer wording, not customer satisfaction. Citation appearances across all captured answers are a different measure from benchmark answers citing a page.

Cited pages and disputed claims still need independent verification. This edition sets a broader baseline; earlier editions used a narrower question set and are not treated as a trend.

Our first three priorities

First, verify the home-loan facts. Ownership, operating status and dates should be confirmed against primary sources before any page brief or publisher request proceeds, and ambiguous entity matches should stay out of the correction queue.

Second, prepare page briefs for the pages that matter most, starting with the home-loans page and the credit-card education pages, using approved facts and the buyer questions observed in the research. Existing coverage should be checked before new content is proposed.

Third, build the buyer-question content around no-fee and online banking comparisons, where Discover already appears in answers. A fresh scorecard at the end will show where things stand, reported separately from the work completed.

This report supports CiteWorks Studio's examination of how AI search is recommending Best Banks. For more detail, you can also read Best Banks: AI Visibility Discovery Index.

/ Learn More

Understanding AI search visibility.

AI search experiences create answers by pulling information from many places online and summarizing it into a single response.

What Is AI Citation Intelligence?
AI citation intelligence is the process of measuring where AI platforms source their information and how frequently a brand is mentioned or referenced in AI-generated responses. Because LLMs synthesize across multiple sources, the sites and brands that appear repeatedly tend to influence how a topic or company is framed. This practice focuses on identifying which sources shape AI outputs and tracking brand visibility across different AI systems.
What Is Citation Architecture?
Citation architecture describes the set of sources that consistently inform how AI systems talk about a brand, product, or topic. LLMs draw from websites, articles, forums, and public discussion, and the sources they rely on most often become the backbone of their answers. Building strong citation architecture means ensuring that accurate, credible, high authority sources are the ones most likely to shape the way AI tools summarize and recommend a brand.
What Is Generative Engine Optimization?
Generative engine optimization (GEO) is the practice of improving the chances that AI systems use and cite your brand or content when generating answers. While traditional SEO is centered on ranking pages in search results, GEO focuses on how LLMs retrieve, interpret, and combine information when responding to a question. The objective is to strengthen the content and sources AI systems rely on, so your brand is treated as a trusted reference in AI responses.
What Is AI Share of Voice?
AI share of voice tracks how often a brand appears in AI-generated answers compared with competitors in the same category. It reflects visibility across AI platforms such as ChatGPT, Gemini, Claude, and Perplexity. Monitoring AI share of voice helps organizations see whether AI systems consistently include and recommend their brand for key queries or whether competitor brands are showing up more often.

About The Author

Mark Huntley

Mark Huntley

Founder and CEO

Mark Huntley, J.D. is founder of CiteWorks Studio, a strategic advisory focused on visibility, authority, and recommendation presence in AI-shaped search environments. His work centers on embedding-level GEO, vector optimization, and cosine gap engineering — helping brands align their digital presence with the retrieval systems that increasingly shape discovery, interpretation, and choice.

Archived editions

Browse earlier published research snapshots for this article. The latest edition remains at the current URL.

VIEW ALL CASE STUDIESREQUEST AN AI VISIBILITY AUDIT