SEO competitor analysis: find gaps in Google and AI answers

Start any SEO competitor analysis with this: your competitors in search are often not the businesses you meet at trade fairs. For many of the searches that matter to you, the pages above yours belong to marketplaces, directories, publishers and Reddit threads, and the sources AI answers cite can be different again. In an updated Ahrefs study reported by Search Engine Journal in 2026, only about 38% of URLs cited in Google's AI Overviews also ranked in the organic top 10 for the same query, across 863,000 keywords and 4 million cited URLs.

So a competitor analysis that only compares keyword rankings now misses a large part of the picture. This guide gives you a free-first workflow that covers both: who ranks, who gets cited in AI answers, what their pages have that yours lack, and which of those gaps are worth your time.

Who this is for: small-business owners and in-house marketers who want to know why competitors outrank them or get named in AI answers more often, without buying an enterprise SEO suite. What you will be able to do: map your real search competitors, compare rankings and AI citations with free tools, score every gap on one sheet, and decide topic by topic whether to compete, differentiate or skip.

What SEO competitor analysis covers in 2026

SEO competitor analysis is the process of finding out which sites win the searches you care about, in Google's results and in AI answers, and working out what they do that you don't. Its output should be a short, prioritised list of gaps to close, not a spreadsheet of competitor metrics.

It used to mean comparing keyword rankings and backlinks. That is still part of the job, but three things changed in 2025 and 2026.

1. Rankings and citations have separated. In the same 2026 Ahrefs data, the other AI Overview citations split almost evenly between positions 11 to 100 (31.2%) and pages beyond position 100 (31.0%). Ahrefs said better citation detection explains part of the fall from 76% in its earlier study, but ranking well clearly no longer guarantees being cited.
2. Google's own AI surfaces disagree. Ahrefs found that AI Mode and AI Overviews cited the same URLs only 13% of the time across 540,000 query pairs (US data from September 2025), even though their answers had an average semantic similarity of 86%. You need to check each surface on its own.
3. AI answers search wider than your keyword. Google says AI Overviews and AI Mode may use a query fan-out technique, issuing several related searches across subtopics to build one answer. A competitor can be cited because it ranks for a sub-question you never targeted.

The same Google page says there are no additional technical requirements for its AI features: a page must be indexed and eligible to show in Search with a snippet. The fundamentals that win rankings also make you eligible for citations, which is why one analysis can cover both.

What you need before you start

• A list of 15 to 30 priority searches: the service terms and questions that bring you customers. Our keyword research guide shows how to build one from real customer language.
• Access to Google Search Console and Bing Webmaster Tools for your own site.
• A spreadsheet with one row per search and a column for each check in this guide.
• Half a day for the first run. Repeats are quicker, because the sheet and the method already exist.

Identify your real search competitors, not just business rivals

Your real search competitors are the domains that appear most often in Google's results and AI answers for your priority searches, whether or not they sell what you sell. Find them by tallying domains across your query list, not by naming the rivals you already know.

A local accountant might name three other firms in town. Search the firm's top queries and the results may be led by a national directory, a government guidance page, a software company's blog and a Reddit thread. Each of those takes clicks and citations that could have gone to the accountant, so each is a competitor for that query.

The four kinds of search competitor

• Direct search rivals sell what you sell and rank for the same searches. These deserve the deepest analysis, because you can realistically beat them.
• Marketplaces and directories (booking platforms, review sites, listing directories) list you and your rivals side by side. You rarely outrank them for broad terms, but you can be listed well on them.
• Publishers and reference sites (news sites, magazines, Wikipedia, government and university pages) win information queries and are cited heavily by AI.
• Communities (Reddit, forums, YouTube, LinkedIn) supply first-hand opinions. In Peec AI's 2026 analysis of 30 million sources, Reddit was the most-cited domain across ChatGPT, AI Mode, Gemini, Perplexity and AI Overviews, with YouTube, LinkedIn, Wikipedia and Forbes completing the top five.

Some of what AI cites is simply out of reach. In Ahrefs' analysis of ChatGPT's 1,000 most-cited pages (September 2025 data), Wikipedia made up 29.7% of the pages, homepages and landing pages 23.8% and educational pages 19.4%, and Ahrefs judged about two-thirds of the list to be off-limits to marketers. 28.3% of those pages had no organic keyword rankings at all. Knowing which competitors you cannot displace stops you wasting months chasing them.

How to map your competitors in an hour

1. Pick 15 to 30 searches from your list. Mix service terms ("emergency plumber Leeds"), cost and comparison questions, and the problems customers describe before they buy.
2. Search each one on Google in a private browser window, from the area you serve. Note the domains in the top 10 organic results, the businesses in the map pack if one appears, and the sources linked in any AI Overview.
3. Ask the same question in AI Mode, ChatGPT search and Perplexity, plus Copilot if your customers use Microsoft products. Note every cited domain and every business named in the answer.
4. Tally the domains. Count how many of your searches each domain appears in, and label it with one of the four types above.
5. Keep a shortlist of five to eight. Analyse in depth the direct rivals that appear most often, plus any marketplace or community that dominates several queries.

Compare rankings and keyword gaps

A competitor keyword gap is a search where a competitor ranks and you don't, or where you rank well below them. Find gaps from two directions: your own Search Console data for searches where you are close, and competitors' sites for topics you have no page for at all.

Start with your own data

Search Console's Performance report lists the queries you already appear for, with clicks, impressions and average position. Filter to queries with an average position between roughly 8 and 20 and a healthy number of impressions. These are searches where Google already sees you as relevant but a competitor sits above you. Look each one up on your tally sheet to see who.

Then find the topics competitors cover and you don't

Topic gaps are often bigger than ranking gaps: a rival may have pages on "boiler service cost" or "combi vs system boiler" while you have nothing that could rank. The free Competitor Content Gap Finder finds these by reading sitemaps. Enter your site and up to two competitors and it lists the topics and formats they cover that you don't, ranking comparisons, cost guides, "best" lists and how-tos higher. Treat the result as candidates, not a to-do list: a topic belongs on your site only if it connects to what you sell.

Paid suites such as Ahrefs and Semrush offer keyword gap reports from their own ranking databases. They are faster on large sites, but their volumes and positions are estimates, so confirm any gap that matters with a live search.

Measure share of voice without a paid tool

SEO share of voice is the share of the available visibility for a fixed set of searches that goes to each competitor. Paid tools weight it by estimated click-through rates; you can approximate it for free with a simple points method.

1. For each search on your list, give each top 10 result points: 10 for position 1, 9 for position 2, down to 1 for position 10.
2. Add up each domain's points across all searches.
3. Divide each domain's total by the points available (55 per search) to get its share.

Example (hypothetical): across 20 searches there are 1,100 points available. If your pages score 132 points, your share of voice is 12%. A rival scoring 286 points holds 26%. Keep the same searches every quarter, and the trend becomes more useful than any single figure. For AI answers, use the share of sampled answers that cite or name each competitor instead.

Find AI citation gaps

An AI citation gap is a question where AI answers cite or name a competitor but not you. Find them by asking each priority question several times on each AI surface, logging every cited URL, and then comparing your page with the page that wins.

Sample, don't spot-check

One check tells you very little. In SparkToro's 2026 research with Gumshoe, an AI-tracking company (a conflict of interest the authors disclose), there was less than a 1-in-100 chance that ChatGPT or Google's AI would return the same list of brands in any two of 100 responses. The researchers ran each prompt 60 to 100 times and suggest tracking how often a brand appears rather than its position: brands that appeared across many runs showed a fairly stable pattern, even though their position in each list was close to random.

Few small businesses can run 100 checks per question by hand. A workable compromise is five runs per question on each surface for your ten most valuable questions, each recorded as cited, named without a link, or absent. That gives you a frequency rather than a rank. A competitor cited in four of five runs has a real lead; one in five may be chance. Paid AI visibility trackers can automate larger samples if the questions justify the cost.

What to log for each answer

Two kinds of citation gap

Free tools cannot show which pages an AI system read before answering, but your rankings give a good clue, and each kind of gap needs a different fix.

• You don't rank for the question or its obvious sub-questions. Your page is probably not in the pool of candidates the AI reads. Treat it as a rankings and coverage problem and go back to the keyword gap.
• You rank on page one, but a rival's page is cited. Your page is probably being read and losing at the passage level. Compare the two pages directly.

The free AI Citation Gap Checker makes that comparison. Enter your page and the cited competitor page, and it measures both the same way, factor by factor (answer-first opening, facts and figures, sources, author, last update, AI crawler access and more), then lists the gaps to close first. Focus on factors where the cited page is clearly ahead, not small differences.

When the cited source isn't a competitor's own site

For "best" and "which" questions, the cited page is often a third-party list, review site or forum thread. That leaves two routes: make your own website a top-cited domain through owned content, or get listed on the domains that are already cited. Paid trackers such as Profound (see its guide to tracking visibility in AI search) report which domains are cited for your prompts. For a small business, the second route usually means accurate directory listings, genuine reviews and a fair mention in independent roundups, never paid placement that isn't disclosed. Our guide to getting cited in AI search covers both routes step by step.

Use the first-party data you already have

Bing Webmaster Tools' AI Performance report, in public preview since February 2026, shows how often your pages are cited in AI answers across Microsoft Copilot, Bing's AI-generated summaries and select partner integrations, with the grounding queries behind those citations. Citations are not clicks. In June 2026 Bing added a Citation Share view: the percentage of citations for a grounding query that go to your site, out of all citations shown for it. Bing calls it an observational metric, not a competitive scoreboard, and it does not reveal competitor domains (checked October 2026). A low share on a query you care about is still worth investigating.

Google's generative AI performance report in Search Console shows impressions for your pages in AI Overviews and AI Mode, by page, country, device and date, but no queries and nothing about competitors (checked October 2026). Use it to see which of your pages already appear. Our guide to the Search Console AI performance report explains how to read it.

Analyse the content patterns that win

Look at the pages that rank and get cited for a topic as a group, and note what they share: format, depth, subtopics and entities covered, evidence and freshness. Then fill the genuine holes in your page and add what none of them has, rather than copying what they all have.

Google's own self-assessment questions frame this comparison well. Its guidance on creating helpful, reliable, people-first content asks "Does the content provide substantial value when compared to other pages in search results?" and "Does the content provide original information, reporting, research, or analysis?" A competitor analysis answers the first question with evidence instead of a guess.

Six things to compare

1. Format. Are the winners service pages, how-to guides, comparisons, price guides, lists or tools? If every result for "X cost" is a price guide and you offer a plain service page, the format is the gap. Our post on service pages versus blog posts helps you choose.
2. Coverage. Which subtopics and questions do most winners answer that you skip?
3. Entities. Which products, standards, places, tools and terms do they name?
4. Evidence. Do they give prices, figures with sources, photos, worked examples or named experts?
5. Answer position. Does each section open with a direct answer, or bury it under an introduction?
6. Freshness. When were they last updated, and do they show the date?

Two free tools speed this up. The Cited Content Pattern Analyzer takes the two to six pages an AI answer cites, classifies each (list, how-to, comparison, pricing guide and so on) and measures length, headings, tables, sources and freshness, so you see the shared pattern and how your page differs. The Entity Coverage Checker compares the entities on your page with up to three competitor pages and shows the ones most of them cover and you don't.

Copy the gaps, not the pages

The temptation after a content comparison is to rewrite your page as an average of the top five. Google's May 2026 guide to optimising for generative AI features warns against exactly that: "Don't just recycle what others on the internet have already said, or could easily be produced by a generative AI model." It contrasts commodity content, based on common knowledge, with non-commodity content that offers expert or first-hand takes that go beyond the ordinary.

So use the comparison in two passes. First, fill the genuine holes: a missing subtopic, a missing price range, a question every winner answers and you don't. Second, add something none of them has: your own process, real local detail, photos of your own work, a decision table or a worked calculation. The second pass is what makes your page worth citing over theirs, and it is the core of our content SEO work.

Compare technical and page experience basics

Check that your priority pages are indexed, readable without JavaScript and reasonably fast, then compare the same basics on the pages that beat you. Technical issues rarely explain a whole competitive gap, but a page that is blocked or not indexed cannot win at all.

The Core Web Vitals thresholds are a Largest Contentful Paint (LCP) of 2.5 seconds or less, an Interaction to Next Paint (INP) of 200 milliseconds or less and a Cumulative Layout Shift (CLS) of 0.1 or less, each measured at the 75th percentile of page views. INP replaced First Input Delay as a Core Web Vital in March 2024. You can test a competitor's URL in PageSpeed Insights exactly as you test your own: its field data comes from the Chrome User Experience Report, and when a page has too little real-user data the report falls back to the whole site's origin, or shows no field data for low-traffic sites.

For your own site, run the free Website SEO Audit: it crawls up to 50 pages the way a search engine does and reports broken links, redirects, duplicate titles, missing meta and speed issues with a health score. Fix anything that stops a priority page being indexed before you work on content; our technical SEO audit checklist goes further. Keep technical gaps in proportion, though: passing Core Web Vitals rarely explains why a far more useful page outranks a thin one.

Compare local signals: profiles, reviews and listings

If you serve a local area, compare your Google Business Profile, reviews and directory listings with the businesses in the map pack and in local AI answers. For local searches these signals often decide who gets named at all.

AI local answers are narrow and inconsistent. In BrightLocal's study of about 200,000 non-branded local prompts, published in September 2026, Google Maps mentioned a tracked business in 66% of searches, against 38% for AI Overviews, 33% for ChatGPT and 32% for AI Mode. AI Mode and AI Overviews shared only 29% of the businesses they named. Google Business Profile made up 28.5% of all citations, used heavily by Google's AI surfaces and hardly at all by ChatGPT. Check each surface separately.

Reviews are where gaps are easiest to see and to close. BrightLocal's Local Consumer Review Survey 2026 of 1,002 US consumers found that 74% look for reviews written in the last three months, 31% will only use a business rated 4.5 stars or higher (up from 17% a year earlier) and 47% will not use a business with fewer than 20 reviews.

Local comparison checklist

Close profile gaps with our Google Business Profile optimisation guide, and close review gaps honestly with the methods in how to get more Google reviews. For how AI Mode, Ask Maps and ChatGPT treat local businesses, read our guide to local SEO for AI search.

Score the gaps: a competitor scorecard template

Put every finding on one scorecard so you can compare topics fairly. Score five areas for each priority topic as green, amber or red against the competitors that beat you, turn the colours into points, and weigh the total against the topic's value and how winnable it is.

How we built this: we turned the checks in the previous five sections into one rubric, with thresholds that rely only on free data: live searches, sampled AI answers, Search Console, PageSpeed Insights and Google Business Profile. It is a prioritisation aid, not a model of how Google ranks pages, so adjust the thresholds to your market. You can copy the tables below straight into a spreadsheet.

The five gap areas

Add the points to get a gap score from 0 to 10. Score the local row only for searches that show a map pack; for other searches the maximum is 8. A higher gap score means more work to catch up.

Value and winnability

• Value (1 to 3): 3 if the search is a direct request for what you sell, 2 if it is a research or comparison question that often comes before a purchase, 1 if it is general information.
• Winnable (1 to 3): 3 if direct rivals of a similar size lead, 2 if the results mix rivals with marketplaces or large publishers, 1 if they are dominated by sources you cannot displace, such as Wikipedia, government sites, national brands or community threads.

Priority = Value × Winnable, which gives a score from 1 to 9. Work through topics from the highest priority down. Where two topics tie, do the one with the lower gap score first, because it is quicker to close.

Compete, differentiate or skip: decide topic by topic

The scorecard tells you how big each gap is; who leads the results tells you what kind of move fits. Compete when the gap is in things you control, differentiate when the leaders have strengths you cannot match soon, and skip when the value is low or the results belong to sources you cannot displace.

Worked example: a home-cleaning company in Bengaluru

Example (hypothetical): a small home-cleaning company in Bengaluru scores four topics after running the steps above. The competitors, positions and scores are invented to show the method.

Sofa cleaning comes first. The company ranks seventh (amber), is never cited (red), has no prices while the cited rival does (red), passes the technical checks (green) and trails slightly on reviews (amber). Every gap is in its control, so it adds real price ranges, explains what changes the price and opens with a one-sentence answer.

The move-out checklist ties with deep cleaning on priority but has the lower gap score, so it goes second: the company expands its short checklist with room-by-room time estimates from its own jobs.

Deep cleaning is valuable, but the marketplace wins the broad term. The company completes its marketplace and directory listings, then builds its own page around what those listings don't show (exact scope, products, typical duration, its own before-and-after photos) and targets narrower searches such as post-renovation cleaning. The hard-water topic gets a few lines on the bathroom-cleaning page instead of its own page.

Turn findings into a prioritised plan

Turn the top of your scorecard into a 90-day plan of three to five topics, each with one owner, one main change and one measure. A short list that gets done beats a long one that doesn't.

1. Fix blockers first. Any red technical score on a priority page, such as a page that is not indexed, a blocked crawler or a stray noindex, comes before everything else, because nothing else works until it is fixed.
2. Take the quick "compete" wins. Start with high-priority topics with low gap scores: add the missing subtopic or price range, rewrite the opening so it answers the question directly, and update the date only when the content really changes.
3. Start one "differentiate" project. Pick the highest-value topic that needs a different angle and plan the original material it requires, such as photos, data, a process write-up or a calculator.
4. Close local gaps continuously. Review requests and profile fixes take weeks to show results, so start them in week one and keep them running.
5. Record the baseline. Save this quarter's scorecard, share-of-voice figures and AI citation counts, so next quarter compares like with like.

What not to do

• Don't copy competitors' pages or chase every gap a tool lists. An average of the top five is commodity content, and a topic a rival covers is not automatically one your customers need from you.
• Don't treat one AI answer as a ranking. Use how often each brand appears across repeated runs.
• Don't buy mentions or rank yourself first on your own "best" list. Google's AI guide says seeking inauthentic mentions "isn't as helpful as it might seem", and readers can usually tell.
• Don't judge success on rankings alone. A topic can gain AI citations and enquiries without moving much in the top 10, so track all the scorecard areas.

If you plan to hand the work to a specialist, our post on how to choose an SEO agency lists the questions to ask before you hire.

Repeat every quarter

Re-run the same analysis every three months, with the same searches, the same competitor shortlist and the same scoring rules. Rankings and AI citations move constantly, so the trend across quarters tells you far more than any single snapshot.

What to re-check each quarter

• Your share of voice and AI citation rate for the same searches and questions.
• New entrants to your shortlist, and cited competitor pages that changed format or were updated.
• Your scorecard colours for each priority topic: which moved from red to amber, and which stayed put.
• Your first-party AI data in Search Console and Bing Webmaster Tools.

Run an extra check outside the cycle after a Google core update, or whenever a priority page loses traffic suddenly. Keep a simple change log alongside the scorecard, noting what you changed and when, so you can connect movement to actions rather than guessing.

Competitor analysis is only useful if it changes what you publish. If you would like help running the first round or acting on the findings, talk to us.