Best Bing Copilot API 2026: 8 Scrapers Compared, 4 Tested

If you track how AI search answers talk about your brand, Bing is part of the job. Microsoft gives you no API for it, so you buy a Bing Copilot API or build a scraper.

This post compares eight Bing Copilot APIs that read Copilot Search in Bing. We ran four of them. For the other four, we used their public docs and say so in every row.

On 8 November 2026, cloro asked Bing AI Mode itself: “Is there an API that returns Bing Copilot Search answers with their citations as JSON?” Its answer named SerpApi and the Apify johnvc actor. Both are in this comparison.

Is there an API for Microsoft Copilot answers, and which page does it read?

Microsoft sells no API for Copilot answers. Every Bing Copilot API is a third-party scraper, and each one reads one of three Microsoft pages that carry the name “Copilot”. Those pages give different answers, with different sources, for the same prompt, so an API that reads one tells you nothing about the others.

1. Copilot Search in Bing, at bing.com/copilotsearch. This search tab writes an answer and cites the pages it read. Most “Bing Copilot API” products read this page. Microsoft is testing the label “AI Mode” for it, and cloro sells it as Bing AI Mode.
2. The Copilot box on a normal Bing results page, at bing.com/search. This is a short AI answer above the organic results. DataForSEO and Olostep read it.
3. The Copilot app, at copilot.microsoft.com. This is Microsoft’s chat assistant. Bright Data, querying.ai, Scrapeless, OpenWeb Ninja and Apify’s own microsoft-copilot-search-scraper read it. cloro no longer scrapes the Copilot app: it stopped on 5 November 2026, because Copilot blocks logged-out sessions.

Microsoft also retired the Bing Search APIs, and they never returned Copilot answers. For what Microsoft does offer, see Microsoft Copilot API: official APIs vs scraping.

Pick the page first, then the API. The main comparison covers Copilot Search, and the other two pages get a short section at the end.

How we tested the Bing Copilot APIs

On 8 November 2026 we sent the same 10 prompts, from the US, to three of the APIs:

1. best project management software for small teams
2. how to set up two-factor authentication on google workspace
3. iphone 17 vs pixel 10 camera
4. best high yield savings accounts
5. how to file taxes as a freelancer
6. ibuprofen vs acetaminophen for back pain
7. best time to visit japan
8. how to find cheap flights to europe
9. solar panels vs a heat pump for lowering energy bills
10. best online courses to learn python

We ran cloro’s sync API with country US, two requests at a time. We ran each Apify actor once, with all 10 prompts as one input, and the lofomachines actor with market en-US. Our two Apify runs cost $0.15 (johnvc) and $0.06 (lofomachines) at Free plan prices.

We measured answers returned, time per answer, sources per answer, and inline citation mapping (can you tell which sentence each source supports?). Time per answer is the request time for cloro, the time between answers in one run for johnvc (the first answer came 25.7 s after the run started, including actor start-up), and the actor’s own durationMs figure for lofomachines.

The SearchApi figures come from a different probe. cloro runs a competitor monitor, a set of scheduled checks that send prompts to other APIs. On 7 November 2026 it sent 10 different prompts to SearchApi’s Bing Copilot engine across eight countries.

This test has limits. It is one run of 10 prompts from the US, and the SearchApi probe used other prompts and countries, so its numbers do not compare one for one with ours. We did not run SerpApi, Searlo, ScrapeBadger or the Apify practicaltools actor, because we had no account for them in this test. Their rows use documented figures only and carry the label “docs”.

Results: 8 Bing Copilot APIs side by side

Vendor	Page it reads	Data	Answers	Time per answer	Sources per answer	Inline citation mapping	Targeting	Price per 1,000 (entry paid plan)	Free tier
cloro	bing.com/copilotsearch	Tested	10 of 10	15.0 to 26.5 s, median 19.4 s	9 to 11 sources, 14 to 39 citation pills	citationPills: one entry per chip and source, with domain	Country (all but 6), US state	$3.20 async (Lite, $30); $1.60 on Hobby	500 credits a month (125 async requests)
SearchApi	bing.com/copilotsearch	Tested (monitor probe, 7 Nov, 8 countries)	10 of 10	11 to 41 s	2 to 7 reference links	reference_indexes on text blocks	36 Bing markets	$4.00 (Developer, $40)	100 requests
SerpApi	bing.com/copilotsearch	Docs	Not run	8.02 to 12.55 s (doc samples)	Not run	reference_indexes on text blocks	City-level location, lat/lon; no country parameter	$25.00 (Starter), $15.00 (Developer)	250 searches a month
Searlo	Copilot Search on Bing	Docs	Not run	10 to 30 s (stated)	Not run	citationPills (position, label, description)	gl and hl, mapped to a Bing market “best effort”	$6.38 (one-time pack)	3,000 credits (375 answers)
ScrapeBadger	bing.com/copilotsearch	Docs	Not run	15 to 25 s (stated)	Not run	Citations with start_index, end_index, matched_text	Country, market, strict_country (30 markets)	$3.00 pay as you go, $1.63 (Starter)	1,000 credits (50 calls)
Apify johnvc	Not stated (output schema matches SerpApi’s)	Tested	10 of 10	16.0 to 33.1 s, median 19.5 s	8 to 11 references	reference_indexes on some blocks; 4 of 10 answers had none	None	$15.00 (Apify Free plan), $12.86 (Gold)	Apify Free plan
Apify lofomachines	Not stated	Tested	10 rows: 3 complete, 7 marked partial	68.9 to 215.5 s, median 206.9 s	2 to 5 sources	13 to 29 inline citations with text anchors	11 markets	$6.00 (Apify Free plan), $4.50 (Gold), plus $0.00005 per start	Apify Free plan
Apify practicaltools	bing.com/copilotsearch	Docs	Not run	Not stated	Not run	Citations (index, url)	Not stated	$100.00	Apify Free plan

Prices are list prices from each vendor’s public page on 8 November 2026. Each vendor section below shows the arithmetic.

Of the four Bing Copilot APIs we tested, three returned a complete answer for every prompt, and lofomachines marked 7 of 10 as partial. cloro and johnvc ran at about the same speed. The bigger differences are in targeting and in how each response ties a source to a sentence. That decides whether you can report “Bing cited our page for this claim” or only “Bing cited our page”.

1. cloro: Bing next to Google AI Mode and ChatGPT on one key

cloro’s Bing Copilot API, sold as Bing AI Mode, reads Copilot Search at bing.com/copilotsearch. The same API key also returns ChatGPT, Gemini, Google AI Mode, AI Overview and Google Search, so one integration covers each engine you report on.

You send a prompt and a country. cloro opens Copilot Search from that location, parses the page and returns JSON.

curl -X POST https://api.cloro.dev/v1/monitor/bing/aimode \
  -H "Authorization: Bearer $CLORO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "best high yield savings accounts", "country": "US", "state": "CA", "include": {"markdown": true, "html": true}}'

The response has five parts:

• result.text: the answer.
• result.sources: the pages Bing read, then any other page a citation chip links to, each with position, label, URL and description.
• result.citationPills: one entry for each chip and source pair, with position, label, URL, domain and pill ID. This is how you tie a source to the part of the answer it supports.
• result.markdown: headings, tables and Bing’s follow-up question. These are only in the markdown.
• result.html: a URL to the rendered page, for audits.

Targeting covers every country except BY, CN, IR, RU, SY and VE, plus a US state. A prompt can be up to 3,500 characters after URL encoding.

An async request costs 4 credits, the same as Google AI Mode. The $30 Lite plan has 37,500 credits, which is 9,375 async requests or $3.20 per 1,000. The $100 Hobby plan (250,000 credits) is $1.60 per 1,000, and larger plans go down to $1.22 per 1,000 on the $5,000 Enterprise plan.

The sync API adds 2 credits, and a US state adds 2. The free plan gives 500 credits a month, which is 125 async requests. For batches, POST /v1/async/task and /v1/async/task/batch with taskType: "BING_AIMODE" return results by webhook or polling, with no sync surcharge.

What we saw: 10 of 10 answers, 15.0 to 26.5 s each (median 19.4 s) on the sync API, 9 to 11 sources and 14 to 39 citation pills per answer.

• Best for: scheduled brand tracking that puts Bing next to Google AI Mode and ChatGPT, and any job that needs a US state.
• Trade-off: below about 30,000 requests a month ScrapeBadger lists a lower total, and from about 500,000 SearchApi does. cloro does not cover the Copilot app at copilot.microsoft.com.

2. SearchApi: market-level targeting on a $40 plan

SearchApi’s Bing Copilot API uses engine=bing_copilot, and the request_url in its docs is bing.com/copilotsearch. You pass q, a market_code (36 values, such as en-us or de-de) and an optional zero_retention flag. There is no city or state parameter.

The response has text_blocks with reference_indexes, a markdown field that ends with a source list, and reference_links (index, title, link, displayed link, snippet, source). It has no follow-up questions.

The API is sync. Its OpenAPI file sets a 90 s timeout, and you can use at most 20% of your plan’s credits per hour. Pricing: 100 free requests, then monthly plans from $40 for 10,000 searches ($4.00 per 1,000) through $100 for 35,000, $250 for 100,000 and $500 for 250,000, up to $1,500 for 1 million ($1.50 per 1,000) and $5,000 for 5 million ($1.00 per 1,000). The page states no Copilot surcharge, and SearchApi charges only for successful searches.

What we saw: in cloro’s monitor probe on 7 November 2026, SearchApi’s Bing Copilot engine answered 10 of 10 prompts across eight countries in 11 to 41 s, with 2 to 7 reference links each. Those were different prompts from the test above.

• Best for: market-level checks in many languages, and the lowest list price per 1,000 at a million requests a month or more.
• Trade-off: no city or state targeting, and the hourly cap of 20% of plan credits limits large bursts.

3. SerpApi: city-level location targeting

SerpApi introduced its Bing Copilot API on 16 June 2026. It reads Copilot Search: the search_metadata.bing_copilot_url field in its docs is a bing.com/copilotsearch URL.

You call GET /search?engine=bing_copilot with q, a city-level location or lat and lon, and options for no_cache, async, zero_trace and output (JSON, HTML or markdown). The docs list no country, market or language parameter.

The JSON has a header, header video, image and video links, text_blocks (paragraph, heading, list, table, code block) and references (index, title, link, snippet, source). Blocks carry reference_indexes, snippet links and highlighted words. There is no markdown field in the JSON (use output=md for that) and no follow-up questions. Sample total_time_taken values in the docs are 8.02 to 12.55 s.

Pricing: 250 free searches a month, Starter $25 for 1,000 ($25.00 per 1,000), Developer $75 for 5,000 ($15.00 per 1,000). The pricing page names no Copilot surcharge and says each request counts as one search, but it does not state this for Copilot specifically.

Not tested.

• Best for: answers as seen from a named city.
• Trade-off: no country or market parameter, and the highest entry price per 1,000 of the vendors here except the practicaltools actor.

4. Searlo: one-time credit packs and follow-up questions

Searlo sells its Bing Copilot API as a Copilot scraper API that reads “Copilot Search on Bing”. You call GET /api/v1/search/ai/copilot with q, gl and hl, and Searlo maps gl and hl to a Bing market on a “best effort” basis. A cache parameter exists, and answers are cached for one hour.

The response has answer, answerMarkdown, citations, citationPills (position, label, description), sources, related (follow-up questions) and a complete flag. The page states 10 to 30 s per answer.

An answer costs 8 credits (a Google search costs 1). Credits come in one-time packs, not a subscription. The Micro pack is $3.99 for 625 answers ($6.38 per 1,000). New accounts get 3,000 free credits, which is 375 answers.

Not tested.

• Best for: one-off research where a small fixed spend matters more than price per 1,000.
• Trade-off: market targeting is “best effort”, and the one-hour cache can return an older answer.

5. ScrapeBadger: character offsets for every citation

ScrapeBadger’s Bing Copilot API, its Copilot scraper, reads bing.com/copilotsearch through /v1/copilot/ask. You set country, market and strict_country, across 30 curated markets. Each citation carries start_index, end_index and matched_text, so you can find the exact span of the answer it supports. The page states 15 to 25 s per answer.

A call costs 20 credits. Pricing: pay as you go at $0.15 per 1,000 credits is $3.00 per 1,000 calls. The $49 Starter plan has 600,000 credits, which is 30,000 calls at $1.63 per 1,000. New accounts get 1,000 free credits (50 calls).

Not tested.

• Best for: citation analysis at the sentence level, and a low monthly spend up to 30,000 calls.
• Trade-off: 30 markets and no city or state targeting.

6. Apify actors: johnvc, lofomachines and practicaltools

Apify is a marketplace of scrapers called actors. Three actors on it return Copilot Search answers, so each works as a Bing Copilot API that Apify prices per result.

johnvc: SerpApi’s output schema plus a brand flag

johnvc/bing-copilot-api takes a query or a list of queries and an optional brandToTrack. It returns answerHeader, markdown, textBlocks, references, headerVideo, imagesLink, videosLink and brandMentioned. Those output keys match SerpApi’s documented Bing Copilot schema one for one, in camelCase.

The actor page does not name the URL it reads, and its FAQ says it has no country or language control. It costs $0.015 per query, which is $15.00 per 1,000 on the Apify Free plan and $12.86 on Gold.

What we saw: 10 of 10 answers. In one run, answers came 16.0 to 33.1 s apart (median 19.5 s), with 8 to 11 references each. Four of the 10 answers had no text block with a reference index, so for those you cannot tie a source to a sentence.

• Best for: a quick US check with a built-in brand flag, and no plan to buy.
• Trade-off: no country or language control, and source-to-sentence mapping was missing on 4 of 10 answers.

lofomachines: text-anchored citations, slow runs

lofomachines/microsoft-copilot-responses-scraper takes a market (11 values) and a list of brands. It returns text (markdown), tables, sources, inlineCitations (id, url, title, text anchor), relatedQuestions and products. The actor page does not name the URL it reads. It costs $6.00 per 1,000 on the Apify Free plan and $4.50 on Gold, plus $0.00005 per actor start.

What we saw: 10 rows. The actor marked 3 as complete and 7 as “partial”, with the note “The response may be incomplete.” Answers took 68.9 to 215.5 s by the actor’s own timer (median 206.9 s). Each had 2 to 5 sources and 13 to 29 inline citations with text anchors.

• Best for: text-anchored citations and related questions in one output, when a run of several minutes is acceptable.
• Trade-off: 7 of 10 answers marked partial in our run, and the page does not say which Microsoft page the actor reads.

practicaltools: Copilot Search without Bing’s markers

practicaltools/bing-ai-copilot reads bing.com/copilotsearch and returns the answer with citations (index and URL), with Bing’s [^1^] markers removed. It costs $0.10 per result ($100 per 1,000). Not tested.

Not Copilot Search: the SERP box and the Copilot app

These products also turn up in searches for a Copilot scraper or a Microsoft Copilot scraper. They read different pages from the eight Bing Copilot APIs above, so their answers, sources and prices do not compare with the results table.

Vendor	Page it reads	Price per 1,000	Source
DataForSEO	Copilot box in a Bing results page (ai_overview item)	$0.60 Standard, $1.20 Priority, $2.00 Live (per results page)	Update, pricing
Olostep	Copilot box in a Bing results page (copilot_answer)	$1.80 (Starter, $9 for 5,000), $0.50 (Standard, $99 for 200,000)	Guide, pricing
Bright Data	Copilot app (copilot.microsoft.com/chats)	$1.50 pay as you go	Product page
querying.ai	Copilot app (page says “copilot.com”)	$2.50 (Basic, $20 for 8,000)	Engine page
Scrapeless	Copilot app (search, smart, chat, reasoning and study modes)	Not verified	Blog post
OpenWeb Ninja	Copilot app	$5.00 (Pro, $25 for 5,000)	API page
Apify microsoft-copilot-search-scraper	Copilot app (chat, reasoning, smart and study modes)	$200 (Apify Free plan), $5 (Bronze)	Actor page

If you track the classic Bing results page, the SERP box products give you the Copilot box and the organic results in one call, at a lower price per page than any Bing Copilot API above. If your audience uses the Copilot app, use an app scraper. Neither tells you what Copilot Search shows for the same prompt.

How to choose a Bing Copilot API by job

One-off research

You want to see what Copilot Search says for a few dozen prompts, once. Searlo’s $3.99 pack (625 answers) or its 375 free answers, cloro’s 125 free async requests a month, and SerpApi’s 250 free searches all cover this without a paid plan. The johnvc actor ran our 10 prompts for $0.15. To build it yourself instead, read how to scrape Bing Copilot.

Scheduled brand tracking across engines

You run the same prompt set every week and report Bing next to Google and ChatGPT. Here a second and third vendor, each with its own schema, key and bill, costs more than a few cents per 1,000. cloro returns Bing AI Mode, Google AI Mode and ChatGPT on one key, with a consistent sources field. For what to do with the data, see Copilot rank tracking, and for how the two AI Modes differ, see Bing AI Mode vs Google AI Mode.

City-level or state-level targeting

SerpApi is the only vendor here with a city-level location parameter, and cloro is the only one with a US state. SearchApi (36 markets) and ScrapeBadger (30 markets) stop at the market level. The johnvc actor has no location control.

Lowest list price at volume

No vendor is cheapest at every volume. This table takes the cheapest published plan that covers each monthly volume, for async Copilot Search requests, from each vendor’s pricing page on 8 November 2026:

Requests a month	cloro	SearchApi	ScrapeBadger
5,000	$30 (Lite)	$40 (Developer)	$15 (pay as you go)
10,000	$100 (Hobby)	$40 (Developer)	$30 (pay as you go) or $49 (Starter)
30,000	$100 (Hobby)	$100 (Production)	$49 (Starter)
60,000	$100 (Hobby)	$250 (BigData)	Above Starter; higher plans not checked
150,000	$250 (Starter)	$500 (Scale)	Not checked
500,000	$1,000 (Business)	$900 (Octo 500K)	Not checked
1,000,000	$2,000 (Enterprise 2,000)	$1,500 (Octo 1M)	Not checked

ScrapeBadger’s pay-as-you-go figures assume its stated 20 credits per Copilot call at $0.15 per 1,000 credits. cloro’s Lite plan covers 9,375 async requests, so 10,000 needs Hobby. Sync calls and a US state each add 2 credits on cloro, which raises its per-request cost by half for each add-on.

From about 500,000 requests a month, SearchApi’s plans cost less than cloro’s, and its $5,000 plan (5 million searches, $1.00 per 1,000) is the lowest list price in this comparison. The cloro credits also pay for the other engines on the same key, which matters if Bing is one of several engines you track.

To check the output on your own prompts, the free plan’s 500 monthly credits cover 125 async Bing AI Mode requests. Plans are on the pricing page.